def createClare(channelB, channelG, channelR):
# create a CLAHE object(Contrast Limited Adaptive Histogram Equalization)
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
cl1 = clahe.apply(channelB)
cv2.imwrite('bluechannel.jpg', cl1)
clahe1 = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
cl2 = clahe.apply(channelG)
cv2.imwrite('greenchanel.jpg', cl2)
clahe2 = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
cl3 = clahe.apply(channelR)
cv2.imwrite('redChannel.jpg', cl3)
result = cv2.merge((cl1, cl2, cl3))
dirname = "afterEqualization"
os.mkdir(dirname)
cv2.imwrite(os.path.join(dirname, "image after equalization.jpg"), result)
return result
def detect(img, cascade, cascade1, cascade_eyes):
clahe = cv2.createCLAHE(clipLimit=5, tileGridSize=(2,2))
img = clahe.apply(img)
clahe = cv2.createCLAHE(clipLimit=5, tileGridSize=(2,2))
img = clahe.apply(img)
best_diff = 1000
imgF = None
for scale in [float(i)/10 for i in range(11, 15)]:
face = cascade.detectMultiScale(img, scaleFactor=scale, minNeighbors=6, minSize=(30, 30))
if len(face) != 1:
continue
x,y,w,h = face[0]
m = max(w,h)
y1 = max(y-m/3, 0)
y2 = min(y + m + m/3, len(img))
x1 = max(x-m/3, 0)
x2 = min(x + m + m/3, len(img[0]))
image = img[y1:y2, x1:x2]
eyes = cascade_eyes.detectMultiScale(image, minNeighbors=3, scaleFactor=scale)
if len(eyes) != 2:
continue
my_eye_right = max(eyes, key=lambda x:x[0])
my_eye_left = min(eyes, key=lambda x:x[0])
dim_r = my_eye_right[2:]
dim_l = my_eye_left[2:]
cel = (my_eye_left[0]+dim_l[0]/2,my_eye_left[1]+dim_l[1]/2)
cer = (my_eye_right[0]+dim_r[0]/2, my_eye_right[1]+dim_r[1]/2)
#cv2.rectangle(image,cer,(cer[0]+5,cer[1]+5),3)
#cv2.rectangle(image,cel,(cel[0]+5,cel[1]+5),3)
prom = (cer[0] + cel[0])/2.0
width = x2-x1
height = y2-y1
width = x2-x1
height = y2-y1
w1 = width*(1.0/3.0)
w2 = width*(2.0/3.0)
diff = abs(cer[0]-w2) + abs(cel[0]-w1)
if diff/float(width) > 0.2 or cel[1] > height/2 or cer[1] > height/2:
continue
print diff, width, w1, w2, abs(cel[0]-w1), abs(cer[0]-w2)
if best_diff > diff:
best_diff = diff
cerF = cer
celF = cel
imgF = image
if imgF == None:
return []
im = Image.fromarray(imgF)
im = CropFace(im, eye_left=celF, eye_right=cerF, offset_pct=(0.2,0.3), dest_sz=(128,128))
image = np.array(im)
return [image]
def inicio(path):
datalist=[('Component',[0,'1st','2nd','3rd']),]
met=fedit(datalist, title="Select Method")
if met==[0]:
name='/comp1.tif'
if met==[1]:
name='/comp2.tif'
if met==[2]:
name='/comp3.tif'
image=np.array(Image.open(path+'/'+name))
clahe(image)
datalist=[('Clip Limit',[0,'None','2','4','6','8','10','12']),]
met=fedit(datalist, title="Select Clip Limit")
if met==[1]: cl=2.0
if met==[2]: cl=4.0
if met==[3]: cl=6.0
if met==[4]: cl=8.0
if met==[5]: cl=10.0
if met==[6]: cl=12.0
image=cv2.createCLAHE(clipLimit=cl, tileGridSize=(4,4)).apply(image.astype(np.uint8))
plt.close()
maps=[m for m in plt.cm.datad if not m.endswith("_r")]
maps=[0]+maps
maps.sort()
print maps
"""colormap(image)
def make_textures_opencv(src_dir, project_dir, image_list, resolution=256):
dst_dir = project_dir + '/Textures/'
if not os.path.exists(dst_dir):
print "Notice: creating texture directory =", dst_dir
os.makedirs(dst_dir)
for image in image_list:
src = src_dir + "/" + image.name
dst = dst_dir + image.name
if not os.path.exists(dst):
src = cv2.imread(src)
height, width = src.shape[:2]
# downscale image first
method = cv2.INTER_AREA # cv2.INTER_AREA
scale = cv2.resize(src, (0,0),
fx=resolution/float(width),
fy=resolution/float(height),
interpolation=method)
# convert to hsv color space
hsv = cv2.cvtColor(scale, cv2.COLOR_BGR2HSV)
hue,sat,val = cv2.split(hsv)
# adaptive histogram equalization on 'value' channel
clahe = cv2.createCLAHE(clipLimit=3.0, tileGridSize=(8,8))
aeq = clahe.apply(val)
# recombine
hsv = cv2.merge((hue,sat,aeq))
# convert back to rgb
result = cv2.cvtColor(hsv, cv2.COLOR_HSV2BGR)
cv2.imwrite(dst, result)
print "Texture %dx%d %s" % (resolution, resolution, dst)
def CLAHE():
img = cv2.imread('test.jpg',0)
# create a CLAHE object (Arguments are optional).
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
cl1 = clahe.apply(img)
cv2.imwrite('clahe_2.jpg',cl1)
def adaptive_histogram_eq(img):
b,g,r = cv2.split(img)
clahe = cv2.createCLAHE(clipLimit=.1, tileGridSize=(8,8))
b = clahe.apply(b)
g = clahe.apply(g)
r = clahe.apply(r)
return cv2.merge([b,g,r])
def find_hottest_points(cv_image):
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(3,3))
#gray = clahe.apply(img)
gray = clahe.apply(cv_image)
gray = cv2.GaussianBlur (gray, (21,21), 0)
min_thresh = cv2.threshold(gray, min_th, 255, cv2.THRESH_BINARY)[1]
max_thresh = cv2.threshold(gray, max_th, 255, cv2.THRESH_BINARY_INV)[1]
thresh = cv2.bitwise_and(min_thresh, max_thresh)
thresh = cv2.dilate(thresh, None, iterations = 2)
(cnts, _) = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
for c in cnts:
if cv2.contourArea(c) > min_area and cv2.contourArea(c) < max_area:
(x,y,w,h) = cv2.boundingRect(c)
# cv2.rectangle(cv_image, (x, y), (x+w, y+h), (0, 255, 0), 2)
cv2.rectangle(cv_image, (x, y), (x+w, y+h), 0, 2)
continue
cv2.imshow("region_detector", cv_image)
cv2.moveWindow("region_detector",900,0)
cv2.imshow("band_threshold_image", thresh)
cv2.moveWindow("band_threshold_image",900,400)
cv2.waitKey(1)
def adjust_zstack(self):
#Removes bckgd fluor from ea. img.
#trying hist eq
self.zsum()
zsum = self.My_Zsum
zstk = np.copy(self.My_Zstack)
####new:####
for frame in range(zstk.shape[2]):
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
cl1 = clahe.apply(zstk[:,:,frame].astype('uint8'))
self.My_Zstack[:,:,frame] = cl1
###end new###
bckgd_fluors=[]
for frame in range(zstk.shape[2]):
if self.weights==None:
self.bckgd_weights(np.copy(zstk[:,:,frame]), zsum, preview=False)
fluor=np.average(zstk[:,:,frame], weights=self.weights)
bckgd_fluors.append(fluor)
bckgd_fluors=np.array(bckgd_fluors)
for frame in range(zstk.shape[2]):
sbd = zstk[:,:,frame]-bckgd_fluors[frame]*np.ones((zstk[:,:,frame].shape)) #subtracted img
clip_im = np.clip(sbd, 0 , sbd)
self.My_Zstack[:,:,frame] = clip_im
def FindWand():
global rval,old_frame,old_gray,p0,mask,color,ig,img,frame
try:
rval, old_frame = cam.read()
cv2.flip(old_frame,1,old_frame)
old_gray = cv2.cvtColor(old_frame,cv2.COLOR_BGR2GRAY)
equalizeHist(old_gray)
old_gray = GaussianBlur(old_gray,(9,9),1.5)
dilate_kernel = np.ones(dilation_params, np.uint8)
old_gray = cv2.dilate(old_gray, dilate_kernel, iterations=1)
clahe = cv2.createCLAHE(clipLimit=3.0, tileGridSize=(8,8))
old_gray = clahe.apply(old_gray)
#TODO: trained image recognition
p0 = cv2.HoughCircles(old_gray,cv2.HOUGH_GRADIENT,3,50,param1=240,param2=8,minRadius=4,maxRadius=15)
if p0 is not None:
p0.shape = (p0.shape[1], 1, p0.shape[2])
p0 = p0[:,:,0:2]
mask = np.zeros_like(old_frame)
ig = [[0] for x in range(20)]
print "finding..."
threading.Timer(3, FindWand).start()
except:
e = sys.exc_info()[1]
print "Error: %s" % e
exit
def claheAdjustImages(self):
#clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
clahe = cv2.createCLAHE(clipLimit=6.0,tileGridSize=(8, 8))
#self.logger.info("The clahe type {}".format(type(clahe.type)))
for i, bgr_img, in enumerate(self.images):
self.logger.info("color adjusting image {}".format(i))
# first convert RGB color image to Lab
lab_image = cv2.cvtColor(bgr_img, cv2.cv.CV_RGB2Lab)
#cv2.imshow("lab_image before clahe", lab_image);
# Extract the L channel
lab_planes = cv2.split(lab_image)
# then apply CLAHE algorithm to L channel
#dst = [] #creating empty array to store L value
#dst = clahe.apply(lab_planes[0]) # L channel = lab_planes[0]
lab_planes[0] = clahe.apply(lab_planes[0])
# Merge the the color planes back into an Lab image
lab_image = cv2.merge(lab_planes,lab_planes[0])
#cv2.imshow("lab_image after clahe", lab_image);
# convert back to RGB space and store the color corrected image
self.images[i] = cv2.cvtColor(lab_image, cv2.cv.CV_Lab2RGB)
# overwrite old image
#self.images[i] = cv2.imwrite(cl1)
#cv2.imshow("image original", bgr_img);
#cv2.imshow("image CLAHE", self.images[i]);
cv2.waitKey()
def experiement_clahe_and_entropy(file=None):
images = image_generator()
if file is not None:
images = add_image_to_image_generator(images, file)
claheizer = cv2.createCLAHE()
for fn, im in images:
image_arrays = []
titles = []
# plot original image
titles.append("Original")
image_arrays.append(im)
# clahe
clahe = cv2.cvtColor(im.astype("uint8"), cv2.COLOR_RGB2Lab)
tmp = clahe.copy()
tmp[:, :, 0] = claheizer.apply(clahe[:, :, 0])
clahe = cv2.cvtColor(tmp.astype("uint8"), cv2.COLOR_Lab2RGB)
titles.append("CLAHE")
image_arrays.append(clahe)
tmp = cv2.cvtColor(clahe, cv2.COLOR_RGB2Lab)
ent = skimage.filters.rank.entropy(tmp[:, :, 0], skimage.morphology.disk(5))
titles.append("Entropy L")
image_arrays.append(ent)
# plot
subplot_images(image_arrays, titles=titles, show_plot=True, suptitle=fn.split("/")[-1])
def colorchange(pic):
img = cv2.imread(pic)
for k in range(n):
i = int(numpy.random.random() * img.shape[1])
j = int(numpy.random.random() * img.shape[0])
if img.ndim == 2:
img[j,i] = 255
elif img.ndim == 3:
img[j,i,0]= 255
img[j,i,1]= 255
img[j,i,2]= 255
#cv2.imwrite("th3.png", img, [int(cv2.IMWRITE_PNG_COMPRESSION), 0])
dst=cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
cl1 = clahe.apply(dst)
#cv2.imwrite("cl1.png", cl1, [int(cv2.IMWRITE_PNG_COMPRESSION), 0])
th4 = cv2.adaptiveThreshold(cl1,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C,\
cv2.THRESH_BINARY,11,2)
#th3 = cv2.adaptiveThreshold(img,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C,\
#cv2.THRESH_BINARY,11,2)
cv2.imwrite("im2.png", th4, [int(cv2.IMWRITE_PNG_COMPRESSION), 0])
cha=cv2.imread('im2.png')
ori = cv2.imread(pic)
cha[numpy.where((cha == [255,255,255]).all(axis = 2))] = [b,g,r]
cha[numpy.where((cha == [0,0,0]).all(axis = 2))] =[B,G,R]
dst= cv2.addWeighted(cha,0.7,ori,0.3,0)
new= cv2.bilateralFilter(dst,7,75,75)
return new
def convert_hsv(self, img_file):
pix = cv2.imread(os.path.dirname(__file__) + img_file)
r_size = 5
c_size = r_size - 1
# using HSV space
hsv = cv2.cvtColor(pix, cv2.COLOR_BGR2HLS)
pix_s = hsv[:, :, 1]
cv2.imshow('pix-s', pix_s)
# Adaptive Threshold -> CLAHE
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(c_size, c_size))
pix_clahe = clahe.apply(pix_s)
pix_bin = cv2.adaptiveThreshold(pix_clahe, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY_INV,
blockSize=r_size*2 + 1, C=r_size)
cv2.imshow('pix-bin', pix_bin)
pix_threshold = histogram_equalization.ostu_algorithm(pix_bin, r_size - 2)
cv2.imshow('pix-ostu', pix_threshold)
# fe = FingerprintEnhance()
# pix_enhance = fe.gabor_filter(pix_threshold)
# cv2.imshow("pix-enhance", np.array(pix_enhance))
gabor = GaborFilter()
pix_gabor = gabor.process(pix_threshold)
cv2.imshow("pix-gabor", np.array(pix_gabor))
cv2.waitKey()
def read_images():
X,y = [], []
cascade_face = cv2.CascadeClassifier()
cascade_face.load("classifiers/lbpcascade_frontalface.xml")
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
#emociones = ("enojado", "feliz", "neutral", "sorpresa", "triste")
#carpeta_emociones = "D:/Respaldo Jose Luis/proyecto RVERK/RafD_Ordenado/"
carpeta_emociones = "D:/Francisco/Pictures/Camera Roll/"
indice = -1
for emocion in emociones:
imagenes = glob.glob(carpeta_emociones + emocion + "\\*.jpg")
indice += 1
for imagen in imagenes:
try:
im = cv2.imread(imagen, cv2.IMREAD_GRAYSCALE)
faces = cascade_face.detectMultiScale(im)
if len(faces) > 0:
for (corX, corY, w, h) in faces[:1]:
im = im[corY:corY + h, corX:corX + w]
#im = cv2.resize(im ,(100, 100))
X.append(np.asarray(im, dtype=np.uint8))
y.append(indice)
except IOError, (errno, strerror):
print "I/O error({0}): {1}".format(errno, strerror)
except:
def CLAHE(filename, clip_limit=2.0): img = cv2.imread(filename, cv2.IMREAD_GRAYSCALE) clahe = cv2.createCLAHE(clipLimit=clip_limit) equalized = clahe.apply(img) return [img, equalized]
def hist_eq(image_dir = 'test_hist/', target_dir = 'test_result_hist/', method = 'CLAHE'):
if not os.path.exists(target_dir):
os.makedirs(target_dir)
#pic_list = os.listdir(image_dir)
pic_list = glob.glob(image_dir+'/*.jpeg')
list_length = len(pic_list)
util.update_progress(0)
for j, image_path in enumerate(pic_list):
img = cv2.imread(image_path,1)
# Use file name only, without .jpeg
image_name = image_path.split('/')[-1][:-5]
b,g,r = cv2.split(img)
if method == 'HE':
cv2.equalizeHist(b,b)
cv2.equalizeHist(g,g)
cv2.equalizeHist(r,r)
else:
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
clahe.apply(g,g)
if not method =='CLAHE_G':
clahe.apply(b,b)
clahe.apply(r,r)
recombined = cv2.merge((b,g,r))
cv2.imwrite(target_dir + image_name + method +'.jpeg', recombined)
util.update_progress(j/list_length)
util.update_progress(1)
def interpolate_bg(img):
bg = img
op = np.random.randint(0,2)
if op: # flip the image
bg = cv2.flip(bg,np.random.randint(-1,2))
op = np.random.randint(0,2)
if op: # perform a rotation
rows, cols = bg.shape
degree = [90,180,360]
d = np.random.randint(0,3)
M = cv2.getRotationMatrix2D((cols/2, rows/2), degree[d], 1)
bg = cv2.warpAffine(bg, M, (cols, rows))
op = np.random.randint(0,4)
if op == 0: # lighten with equalizer
bg = cv2.equalizeHist(bg)
elif op == 1: # ligten with CLAHE
clahe = cv2.createCLAHE(clipLimit=1.2, tileGridSize=(20,20))
bg = clahe.apply(bg)
elif op ==2: # darken the image
bg = contrast_wire(bg, 0.9, 1.10)
else: # do nothing
bg = bg
#cv2.imshow('new iamge',res)
# cv2.imshow('clahe', bg)
# cv2.waitKey(0)
# cv2.destroyAllWindows()
return bg
def capture():
global servo_move_flag
print 'start collecting frame'
faces_detected = []
stream = io.BytesIO()
cam.capture(stream, format='jpeg')
data = np.fromstring(stream.getvalue(), dtype=np.uint8)
frame = cv2.imdecode(data, 1)
#frame = cv2.imread('test/yuchen.jpg')
gray= cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
clear_gray = clahe.apply(gray)
small_gray = cv2.resize(clear_gray,(320,240))
print 'trying to detect faces'
faces = face_cascade.detectMultiScale(small_gray, 1.3, 5)
if faces is not ():
face_cnt = 1
servo_move_flag = 0
for face in os.listdir('faces'):
os.remove('faces/'+face)
for x,y,w,h in faces:
x,y,w,h = wider_area(divider*x,divider*y,divider*w,divider*h)
face_filename = 'face_'+str(face_cnt)
cv2.imwrite('face_buff.jpg',frame[y:y+h,x:x+w])
face_cnt =face_cnt + 1
face = gray[y:y+h,x:x+w]
faces_detected.append(np.asarray(resize(face), dtype=np.uint8))
print 'face detected has been appended'
else:
servo_move_flag = 1
print 'no face detected'
return faces_detected
def ProcessImageHistogram(paramterers):
# Get process realted informations
current = multiprocessing.current_process()
proc_name = current.name
# We are starting
logging.info('[%s] Starting ...' % (proc_name))
# Parameters
input_file = paramterers[0]
output_file = paramterers[1]
contrast_limit = paramterers[2]
# Opening file
logging.info('[%s] Opening: %s file ...' % (proc_name, input_file))
img = cv2.imread(input_file, cv2.IMREAD_COLOR)
# We got problem with the opened file
if img is None:
logging.error('[%s] Error opening: %s file!' % (proc_name, input_file))
return -1
else:
# Color image should be splitted to BGR color channels
logging.info('[%s] Split BGR image color channels of: %s file ...' % (proc_name, input_file))
b, g, r = cv2.split(img)
logging.info('[%s] Processing file: %s ... Contrast limit is: %s.' % (proc_name, input_file, contrast_limit))
# create a CLAHE object - Contrast Limited Adaptive Histogram Equalization an apply on every channel
clahe = cv2.createCLAHE(clipLimit=contrast_limit, tileGridSize=(8, 8))
b = clahe.apply(b)
g = clahe.apply(g)
r = clahe.apply(r)
# Merge image color channels
logging.info('[%s] Merge BGR image color channels of: %s file ...' % (proc_name, input_file))
cl1 = cv2.merge((b, g, r))
logging.info('[%s] File has been processed: %s.' % (proc_name, input_file))
# Write image, file type decided based on its extension
cv2.imwrite(output_file, cl1)
logging.info('[%s] File has been written: %s.' % (proc_name, output_file))
return 0
def histogram_equalization(image, adaptive=False):
if adaptive:
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
image = clahe.apply(image)
else:
image = cv2.equalizeHist(image)
return image
def get_landmarks(lena_gray, x, y, width, height):
keypoints = None
p0 = None
face = face_classifier.detectMultiScale(lena, 1.2, 5)
if len(face) == 0:
return p0,False
# Histogram equalization to improve contrast (CLAHE)
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
lena_gray = clahe.apply(lena_gray)
localizer = Flandmark()
keypoints = localizer.locate(lena_gray, y, x, height, width)
keypoints = np.fliplr(keypoints)
i = 0
p0_len = keypoints.shape[0]
main = np.array([[[]]])
while i < p0_len:
to_add = keypoints[i]
if i == 0:
main = to_add[None,:][None,:]
else:
main = np.concatenate((main, to_add[None,:][None,:]))
i = i + 1
main = np.array(main, dtype='f')
p0 = main
return p0, True
def __init__(self, features=150, gauss_blur=3, med_blur=5, history=1, clip_limit=100.0, tile_grid=(8, 8)):
self.num_features = 150
self.orb = cv2.ORB_create(features)
self.bsb = cv2.createBackgroundSubtractorMOG2(history)
self.clahe = cv2.createCLAHE(clipLimit=clip_limit, tileGridSize=tile_grid)
self.gaussian_blur_factor = gauss_blur
self.median_blur_factor = med_blur
def GenericFilter(self,infile,outfile):
#print infile,":",outfile
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
kernel1 = np.ones((5,5),np.uint8)
kernel2 = np.ones((3,3),np.uint8)
img = cv2.imread(str(infile),0)
cv2.imshow("img",img)
blur=cv2.medianBlur(img,5)
cl1 = clahe.apply(blur)
circles_mask = cv2.dilate(cl1,kernel1,iterations = 6)
circles_mask = (255-circles_mask)
circles_mask = cv2.threshold(circles_mask, 0, 255, cv2.THRESH_BINARY)[1]
edges = cv2.Canny(cl1,100,200)
dilation = cv2.dilate(edges,kernel1,iterations = 1)
display = cv2.bitwise_and(img,img,mask=dilation)
cl2 = clahe.apply(display)
cl2 = clahe.apply(cl2)
ret,th = cv2.threshold(cl2,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
th = 255 - th
thg = cv2.adaptiveThreshold(display,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C,\
cv2.THRESH_BINARY,11,2)
saveimgfile=outfile.split(".")
print saveimgfile[1]
final = cv2.bitwise_and(dilation,dilation,mask=th)
cv2.imwrite("Filtered-images/"+str(saveimgfile[0])+"1."+str(saveimgfile[1]),final)
finalg = cv2.bitwise_and(dilation,dilation,mask=thg)
cv2.imwrite("Filtered-images/"+str(saveimgfile[0])+"2."+str(saveimgfile[1]),finalg)
finalg = 255 - finalg
abso = cv2.bitwise_and(dilation,dilation,mask=finalg)
cv2.imwrite("Filtered-images/"+str(saveimgfile[0])+"orig."+str(saveimgfile[1]),abso)
cv2.waitKey(0)
def capture(root):
frame = queue.get()
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(4,4))
clear_grey = clahe.apply(gray)
faces = face_cascade.detectMultiScale(clear_grey, 1.3, 5)
if faces is not []:
for x,y,w,h in faces:
#print x,y,w,h
x,y,w,h = wider_area(x,y,w,h)
#print x,y,w,h
frame = frame[y:y+h,x:x+w]
clear_grey = clear_grey[y:y+h,x:x+w]
im = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
a = Image.fromarray(im)
b = ImageTk.PhotoImage(image=a)
image_label.configure(image=b)
image_label._image_cache = b # avoid garbage collection
root.update()
cv2.imwrite('temp.jpg',frame)
#save_image_process.terminate
#save_image_process.start()
save_image(root)
def face_detection (cv_image):
cv_image = cv2.flip(cv_image, 1)
# gray_vid = cv2.cvtColor(cv_image, cv2.COLOR_BGR2GRAY)
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
#gray_vid = clahe.apply(gray_vid)
gray_vid = clahe.apply(cv_image)
faces = face_cascade.detectMultiScale(gray_vid, 1.2, 5)
print "entro a funcion"
print faces
for (x, y, w, h) in faces:
cv2.rectangle(cv_image, (x,y), (x+w, y+h), (255, 0, 0), 2)
roi_gray = gray_vid[y:y+h, x:x+w]
roi_color = cv_image[y:y+h, x:x+w]
# eyes = eye_cascade.detectMultiScale(roi_gray)
# print "entro a caras"
# for(ex, ey, ew, eh) in eyes:
# cv2.rectangle(roi_color, (ex, ey), (ex+ew, ey+eh), (0, 255, 0), 2)
# print "entro a ojos"
#cv2.imshow("video_stream", cv_image)
#cv2.waitKey(1)
return cv_image
def load_gray(self):
# print "Loading " + self.image_file
try:
rgb = cv2.imread(self.image_file)
if self.height == 0 or self.width == 0:
self.height, self.width, self.fulld = img_rgb.shape
gray = cv2.cvtColor(rgb, cv2.COLOR_BGR2GRAY)
# cv2.imshow('rgb', rgb)
# cv2.imshow('grayscale', gray)
# histogram equalization
# eq = cv2.equalizeHist(gray)
# cv2.imshow('history equalization', eq)
# adaptive histogram equilization (block by block)
clahe = cv2.createCLAHE(clipLimit=3.0, tileGridSize=(8, 8))
aeq = clahe.apply(gray)
# cv2.imshow('adaptive history equalization', aeq)
# print 'waiting for keyboard input...'
# key = cv2.waitKey() & 0xff
return aeq
except:
print self.image_file + ":\n" + " load error: " + str(sys.exc_info()[1])
def claheAdjustImagesBW(img):
# --> This method does clahe on lab space to keep the color
# transform to lab color space and conduct clahe filtering on l channel then merge and transform back
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
#clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
clahe = cv2.createCLAHE(clipLimit=6.0,tileGridSize=(8, 8))
#print 'converted image to Lab space'
#lab_planes = cv2.split(lab_image)
#lab_planes[0] = clahe.apply(lab_planes[0])
gray = clahe.apply(gray)
print 'apply clahe to image channel 0 --> L sapce'
cv2.imshow("grayCLAHE", gray)
img = cv2.cvtColor(gray, cv2.COLOR_GRAY2BGR)
cv2.imshow("imgCLAHE", img)
# Merge the the color planes back into an Lab image
#lab_image = cv2.merge(lab_planes, lab_planes[0])
print 'merge channels back and transform back to rgb space'
return img
def CenterThreshold(img):
clahe = cv2.createCLAHE(clipLimit=4, tileGridSize=(4,4))
dst = clahe.apply(img[650:800, 1125:1275])
re, dst = cv2.threshold(dst, 253, 255, cv2.THRESH_BINARY_INV)
dst = cv2.GaussianBlur(dst, (11,11),0 )
re, dst = cv2.threshold(dst, 253, 255, cv2.THRESH_BINARY)
return dst
def histogram_equalization(images, adaptive=True):
_images = np.array(images * 255, dtype = np.uint8)
pool = ThreadPool(4)
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
def process_image(image):
#print image.shape, image.dtype
image = image.transpose(1,2,0)
if adaptive:
image[:,:,0] = clahe.apply(image[:,:,0])
image[:,:,1] = clahe.apply(image[:,:,1])
image[:,:,2] = clahe.apply(image[:,:,2])
else:
image[:,:,0] = cv2.equalizeHist(image[:,:,0])
image[:,:,1] = cv2.equalizeHist(image[:,:,1])
image[:,:,2] = cv2.equalizeHist(image[:,:,2])
image = image.transpose(2,0,1)
return image
equalized = pool.map(process_image, _images)
equalized = np.array(equalized, dtype=np.float32)/255.
#visualize_data(np.append(images[:8],equalized[:8],axis=0).transpose(0,2,3,1))
return equalized
def _queryFrame(self):
frame_1 = self.mmc.getLastImage()
#self.npFrame = copy.deepcopy(frame_1);
self.rawFrame = copy.deepcopy(frame_1)
if self.equalizationOn:
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
cl1 = clahe.apply(frame_1)
frame_1 = cl1
else:
brightness_temp = np.array(frame_1, dtype = np.int32)
brightness_temp = brightness_temp + self.BrightnessOffset
clipped = np.clip(brightness_temp, 0, 255)
frame_1 = np.array(clipped, dtype = np.uint8)
frame_flip = cv2.flip(frame_1, 0)
frame_1 = frame_flip
self.npFrame = frame_1;
#frame = [frame_1, frame_1, frame_1]
#print np.var(frame_1)
frame = cv2.cvtColor(frame_1, cv2.COLOR_GRAY2RGB)
#self.npFrame = frame
bitmap = cv2.cv.CreateImageHeader((frame.shape[1], frame.shape[0]), cv2.cv.IPL_DEPTH_8U, 3)
cv2.cv.SetData(bitmap, frame.tostring(), frame.dtype.itemsize * 3 * frame.shape[1])
#frame = OpenCVQImage(frame_3)
self.newFrame.emit(bitmap)
def clahe(image):
gray = cv.cvtColor(image,cv.COLOR_BGR2GRAY)
clahe = cv.createCLAHE(clipLimit=2.0,tileGridSize=(8,8))
dst = clahe.apply(gray)
cv.imshow("clash",dst)
import cv2
import numpy as np
import dlib
from sklearn.svm import SVC
import glob
import random
import math
import itertools
from sklearn.externals import joblib
import os
img_path = '/var/www/html/emorecog/data/test_images/test1.jpg' # THE PATH OF THE IMAGE TO BE ANALYZED
font = cv2.FONT_HERSHEY_DUPLEX
emotions = ["anger", "happy", "sadness"] #Emotion list
clahe = cv2.createCLAHE(clipLimit=2.0,
tileGridSize=(8, 8)) # Histogram equalization object
face_det = dlib.get_frontal_face_detector()
land_pred = dlib.shape_predictor(
"/var/www/html/emorecog/data/DlibPredictor/shape_predictor_68_face_landmarks.dat"
)
def crop_face(i_path):
image = cv2.imread(i_path)
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
dest_i = '/var/www/html/emorecog/data/cap_image/test.png'
# Loading all the HAAR Cascade classifiers
face1 = cv2.CascadeClassifier(
"/var/www/html/emorecog/data/HAARCascades/haarcascade_frontalface_default.xml"
)
no_of_eyes = 0
clean_up()
os.chdir('All_images')
files = glob.glob('*')
for file in files:
try:
img = cv2.imread(file)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
clahe = cv2.createCLAHE(clipLimit=1.6, tileGridSize=(3, 3))
gray = clahe.apply(gray)
faces = face_cascade.detectMultiScale(gray, 1.3, 5)
# print file
smile_found = 0
face_found = 0
eyes_found = 0
img = cv2.imread(file)
for (x1, y1, w1, h1) in faces:
face_found = 1
cv2.rectangle(img, (x1, y1), (x1 + w1, y1 + h1), (255, 0, 0), 2)
def extract_bv(image):
b, green_fundus, r = cv2.split(image)
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
contrast_enhanced_green_fundus = clahe.apply(green_fundus)
# applying alternate sequential filtering (3 times closing opening)
r1 = cv2.morphologyEx(contrast_enhanced_green_fundus,
cv2.MORPH_OPEN,
cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (5, 5)),
iterations=1)
R1 = cv2.morphologyEx(r1,
cv2.MORPH_CLOSE,
cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (5, 5)),
iterations=1)
r2 = cv2.morphologyEx(R1,
cv2.MORPH_OPEN,
cv2.getStructuringElement(cv2.MORPH_ELLIPSE,
(11, 11)),
iterations=1)
R2 = cv2.morphologyEx(r2,
cv2.MORPH_CLOSE,
cv2.getStructuringElement(cv2.MORPH_ELLIPSE,
(11, 11)),
iterations=1)
r3 = cv2.morphologyEx(R2,
cv2.MORPH_OPEN,
cv2.getStructuringElement(cv2.MORPH_ELLIPSE,
(23, 23)),
iterations=1)
R3 = cv2.morphologyEx(r3,
cv2.MORPH_CLOSE,
cv2.getStructuringElement(cv2.MORPH_ELLIPSE,
(23, 23)),
iterations=1)
f4 = cv2.subtract(R3, contrast_enhanced_green_fundus)
f5 = clahe.apply(f4)
# removing very small contours through area parameter noise removal
ret, f6 = cv2.threshold(f5, 15, 255, cv2.THRESH_BINARY)
mask = np.ones(f5.shape[:2], dtype="uint8") * 255
im2, contours, hierarchy = cv2.findContours(f6.copy(), cv2.RETR_LIST,
cv2.CHAIN_APPROX_SIMPLE)
for cnt in contours:
if cv2.contourArea(cnt) <= 200:
cv2.drawContours(mask, [cnt], -1, 0, -1)
im = cv2.bitwise_and(f5, f5, mask=mask)
ret, fin = cv2.threshold(im, 15, 255, cv2.THRESH_BINARY_INV)
newfin = cv2.erode(fin,
cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (3, 3)),
iterations=1)
# removing blobs of unwanted bigger chunks taking in consideration they are not straight lines like blood
#vessels and also in an interval of area
fundus_eroded = cv2.bitwise_not(newfin)
xmask = np.ones(fundus.shape[:2], dtype="uint8") * 255
x1, xcontours, xhierarchy = cv2.findContours(fundus_eroded.copy(),
cv2.RETR_LIST,
cv2.CHAIN_APPROX_SIMPLE)
for cnt in xcontours:
shape = "unidentified"
peri = cv2.arcLength(cnt, True)
approx = cv2.approxPolyDP(cnt, 0.04 * peri, False)
if len(approx) > 4 and cv2.contourArea(
cnt) <= 3000 and cv2.contourArea(cnt) >= 100:
shape = "circle"
else:
shape = "veins"
if (shape == "circle"):
cv2.drawContours(xmask, [cnt], -1, 0, -1)
finimage = cv2.bitwise_and(fundus_eroded, fundus_eroded, mask=xmask)
# blood_vessels = cv2.bitwise_not(finimage)
blood_vessels = finimage
return blood_vessels
def Clahe(frame):
a = cv2.split(cv2.cvtColor(frame, cv2.COLOR_BGR2LAB))
clahe = cv2.createCLAHE(clipLimit=4.5, tileGridSize=(7, 7))
a[0] = clahe.apply(a[0])
return cv2.cvtColor(cv2.merge(a), cv2.COLOR_LAB2BGR)
def convert_object(file_path, screen_size=None, new_file_suffix="scanned"):
""" Identifies 4 corners and does four point transformation """
debug = True if log.level == logging.DEBUG else False
image = cv2.imread(str(file_path))
# convert the image to grayscale, blur it, and find edges
# in the image
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
gray = cv2.bilateralFilter(
gray, 11, 17, 17
) # 11 //TODO 11 FRO OFFLINE MAY NEED TO TUNE TO 5 FOR ONLINE
gray = cv2.medianBlur(gray, 5)
edged = cv2.Canny(gray, 30, 400)
if debug:
previewImage("Edged Image", edged)
# find contours in the edged image, keep only the largest
# ones, and initialize our screen contour
contours, hierarcy = cv2.findContours(
edged, cv2.RETR_LIST, cv2.CHAIN_APPROX_NONE
)
log.debug("Contours found: %s", len(contours))
# approximate the contour
ContourArea = namedtuple('ContourArea', ['curve', 'area'])
contourAreas = [ContourArea(curve=x, area=cv2.contourArea(x))
for x in contours]
contourAreas = sorted(contourAreas, key=attrgetter('area'))
if debug:
previewContours(image, [x.curve for x in contourAreas])
screens = [] # 4 point polygons, repressenting possible screens (rectangles)
for contour in contourAreas:
peri = cv2.arcLength(contour.curve, True)
polygon_less_vertices = cv2.approxPolyDP(contour.curve,
epsilon=0.02 * peri, # approximation accuracy
closed=True)
num_vertices = len(polygon_less_vertices)
if num_vertices == 4:
(x, y, width, height) = cv2.boundingRect(contour.curve)
log.debug(f'x={x} y={y} width={width} height={height}')
screens.append(Screen(fourpoints=polygon_less_vertices, width=width))
if debug:
log.debug(f"Screens found {len(screens)}: {screens}")
previewContours(image, [x.fourpoints for x in screens])
# find largest screen
largest_screen = max(screens, key=attrgetter('width'))
if debug:
previewContours(image, [largest_screen.fourpoints])
# now that we have our screen contour, we need to determine
# the top-left, top-right, bottom-right, and bottom-left
# points so that we can later warp the image -- we'll start
# by reshaping our contour to be our finals and initializing
# our output rectangle in top-left, top-right, bottom-right,
# and bottom-left order
pts = largest_screen.fourpoints.reshape(4, 2)
log.debug("Found bill rectagle at %s", pts)
rect = order_points(pts)
log.debug(rect)
warped = transform_to_four_points(image, pts)
# convert the warped image to grayscale and then adjust
# the intensity of the pixels to have minimum and maximum
# values of 0 and 255, respectively
warp = cv2.cvtColor(warped, cv2.COLOR_BGR2GRAY)
# Replacement for `skimage.exposure.rescale_intensity`
# Contrast Limited Adaptive Histogram Equalization
clahe = cv2.createCLAHE(clipLimit=1.0, tileGridSize=(8, 8))
warp = clahe.apply(warp)
# show the original and warped images
if debug:
previewImage("Original", image)
previewImage("warp", warp)
warp_file = str(file_path.parent / f"{file_path.stem}-{new_file_suffix}.jpg")
cv2.imwrite(warp_file, warp)
log.debug(f"Result: {warp_file}")
if screen_size:
return cv2.cvtColor(
cv2.resize(warp, screen_size), cv2.COLOR_GRAY2RGB
)
else:
return cv2.cvtColor(warp, cv2.COLOR_GRAY2RGB)
from tqdm import tqdm
from utils.general import xyxy2xywh, xywh2xyxy, xywhn2xyxy, clean_str
from utils.torch_utils import torch_distributed_zero_first
import albumentations as A
# Parameters
help_url = 'https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data'
img_formats = ['bmp', 'jpg', 'jpeg', 'png', 'tif', 'tiff',
'dng'] # acceptable image suffixes
vid_formats = ['mov', 'avi', 'mp4', 'mpg', 'mpeg', 'm4v', 'wmv',
'mkv'] # acceptable video suffixes
logger = logging.getLogger(__name__)
CLAHE = False
clahe2 = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
clahe4 = cv2.createCLAHE(clipLimit=4.0, tileGridSize=(8, 8))
# Get orientation exif tag
for orientation in ExifTags.TAGS.keys():
if ExifTags.TAGS[orientation] == 'Orientation':
break
train_transforms = A.Compose([
A.OneOf([
A.CLAHE(clip_limit=2),
A.IAASharpen(),
A.IAAEmboss(),
A.RandomBrightnessContrast(),
],
p=0.35),
gray_img = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# perform contrast limited adaptive equalization
# based on example at:
# http://docs.opencv.org/3.0-beta/doc/py_tutorials/py_imgproc/py_histograms/py_histogram_equalization/py_histogram_equalization.html
# get parameters from track bars
clip_limit = cv2.getTrackbarPos("clip limit", window_name4)
tile_size = cv2.getTrackbarPos("tile size", window_name4)
# perform filtering
clahe = cv2.createCLAHE(clipLimit=clip_limit,
tileGridSize=(tile_size,
tile_size)) # create filter
output = clahe.apply(gray_img) # apply filter
# display image
cv2.imshow(window_name1, gray_img)
cv2.imshow(
window_name2,
hist_lines(cv2.calcHist([gray_img], [0], None, [256], [0, 256])))
cv2.imshow(window_name3, output)
cv2.imshow(
window_name4,
hist_lines(cv2.calcHist([output], [0], None, [256], [0, 256])))
# start the event loop - essential
def main():
cv2.namedWindow('Gamma Correction',cv2.WINDOW_NORMAL)
cv2.namedWindow('Correction',cv2.WINDOW_NORMAL)
cv2.namedWindow('Auto White Balance',cv2.WINDOW_NORMAL)
cv2.namedWindow('Clahe',cv2.WINDOW_NORMAL)
cv2.namedWindow('Clahe LAB',cv2.WINDOW_NORMAL)
cv2.namedWindow('Gaussian',cv2.WINDOW_NORMAL)
cv2.namedWindow('Anisotropic Diffusion',cv2.WINDOW_NORMAL)
cv2.namedWindow('Median Blur',cv2.WINDOW_NORMAL)
cv2.namedWindow('Median Blur 2',cv2.WINDOW_NORMAL)
cv2.namedWindow('YCrCb Median Blur',cv2.WINDOW_NORMAL)
cv2.namedWindow('Output',cv2.WINDOW_NORMAL)
cv2.namedWindow('Thresh Output',cv2.WINDOW_NORMAL)
cv2.namedWindow('Video',cv2.WINDOW_NORMAL)
cv2.resizeWindow('Video',500,500)
cv2.resizeWindow('Clahe',500,500)
cv2.resizeWindow('Clahe LAB',500,500)
cv2.resizeWindow('Gaussian',500,500)
cv2.resizeWindow('Anisotropic Diffusion',500,500)
cv2.resizeWindow('Median Blur',500,500)
cv2.resizeWindow('Median Blur 2',500,500)
cv2.resizeWindow('YCrCb Median Blur',500,500)
cv2.resizeWindow('Output',500,500)
cv2.resizeWindow('Thresh Output',500,500)
imgpath = "/home/charmi/Desktop/Trident_Work/OpenCV/Output/SavedVideos2/Output15.avi"
'''
a ='/home/charmi/Desktop/Trident_Work/OpenCV/Output/SavedVideos/OutputThresh'
b =1
c = '.avi'
filename = a+str(b)+c
#d ='/home/charmi/Desktop/Trident_Work/OpenCV/Output/SavedVideos/Output'
#filename2=d+str(b)+c
while(os.path.exists(filename)):
b+=1
filename = a+str(b)+c
#filename2=d+str(b)+c
codec = cv2.VideoWriter_fourcc('W', 'M', 'V', '2')
framerate = 10
resolution = (640, 480)
VideoFileOutput = cv2.VideoWriter(filename, codec, framerate, resolution)
#VideoFileOutput2 = cv2.VideoWriter(filename2, codec, framerate, resolution)
'''
cv2.createTrackbar('+ve Gamma','Gamma Correction',1,20,emptyFunction)
cv2.createTrackbar('-ve Gamma','Gamma Correction',0,20,emptyFunction)
cv2.createTrackbar('Hue','Correction',1,200,emptyFunction)
cv2.createTrackbar('Saturation','Correction',1,200,emptyFunction)
cv2.createTrackbar('Value','Correction',1,200,emptyFunction)
cv2.createTrackbar('Low H','Thresh Output',0,180,emptyFunction)
cv2.createTrackbar('Low S','Thresh Output',0,255,emptyFunction)
cv2.createTrackbar('Low V','Thresh Output',0,255,emptyFunction)
cv2.createTrackbar('High H','Thresh Output',0,360,emptyFunction)
cv2.createTrackbar('High S','Thresh Output',0,255,emptyFunction)
cv2.createTrackbar('High V','Thresh Output',0,255,emptyFunction)
cv2.createTrackbar('Kernel Size','YCrCb Median Blur',3,15,emptyFunction)
cv2.createTrackbar('Laplacian ksize','Output',3,15,emptyFunction)
cv2.createTrackbar('Clip Limit (Blue)','Clahe',0,100,emptyFunction)
cv2.createTrackbar('Tile Grid Size (Blue)','Clahe',1,20,emptyFunction)
cv2.createTrackbar('Clip Limit (Green)','Clahe',0,100,emptyFunction)
cv2.createTrackbar('Tile Grid Size (Green)','Clahe',1,20,emptyFunction)
cv2.createTrackbar('Clip Limit (Red)','Clahe',0,100,emptyFunction)
cv2.createTrackbar('Tile Grid Size (Red)','Clahe',1,20,emptyFunction)
cv2.createTrackbar('Pair','Clahe',0,1,emptyFunction)
cv2.createTrackbar('Kernel Size','Gaussian',3,15,emptyFunction)
cv2.createTrackbar('Standard Deviation','Gaussian',0,50,emptyFunction)
cv2.createTrackbar('+ve Alpha','Gaussian',0,20,emptyFunction)
cv2.createTrackbar('+ve Beta','Gaussian',0,20,emptyFunction)
cv2.createTrackbar('-ve Alpha','Gaussian',0,20,emptyFunction)
cv2.createTrackbar('-ve Beta','Gaussian',0,20,emptyFunction)
cv2.createTrackbar('Alpha','Anisotropic Diffusion',0,10,emptyFunction)
cv2.createTrackbar('Sensitivity','Anisotropic Diffusion',0,500,emptyFunction)
cv2.createTrackbar('Iterations','Anisotropic Diffusion',1,500,emptyFunction)
cv2.createTrackbar('Kernel','Median Blur',3,15,emptyFunction)
cv2.createTrackbar('Kernel','Median Blur 2',3,25,emptyFunction)
cv2.createTrackbar('FGauss','Gaussian',0,1,emptyFunction)
cv2.createTrackbar('FGamma','Gamma Correction',0,1,emptyFunction)
cv2.createTrackbar('FAWBal','Auto White Balance',0,1,emptyFunction)
cv2.createTrackbar('FAni','Anisotropic Diffusion',0,1,emptyFunction)
cv2.createTrackbar('FB','Median Blur',0,1,emptyFunction)
cv2.createTrackbar('FG','Median Blur',0,1,emptyFunction)
cv2.createTrackbar('FR','Median Blur',0,1,emptyFunction)
cv2.createTrackbar('FMed','Median Blur 2',0,1,emptyFunction)
cv2.createTrackbar('FL','Clahe LAB',0,1,emptyFunction)
cv2.createTrackbar('FA','Clahe LAB',0,1,emptyFunction)
cv2.createTrackbar('FB','Clahe LAB',0,1,emptyFunction)
cv2.createTrackbar('FB','Clahe',0,1,emptyFunction)
cv2.createTrackbar('FG','Clahe',0,1,emptyFunction)
cv2.createTrackbar('FR','Clahe',0,1,emptyFunction)
cv2.createTrackbar('FL','Output',0,1,emptyFunction)
cv2.createTrackbar('FM','YCrCb Median Blur',0,1,emptyFunction)
cap = cv2.VideoCapture(imgpath)
ret, image=cap.read()
(h, w, c) = image.shape
cv2.circle(image, (w//2, h//2), 7, (255, 255, 255), -1)
width2 = float(w)/2
cv2.setTrackbarPos('+ve Gamma','Gamma Correction',16)
cv2.setTrackbarPos('-ve Gamma','Gamma Correction',0)
cv2.setTrackbarPos('Hue','Correction',100)
cv2.setTrackbarPos('Saturation','Correction',100)
cv2.setTrackbarPos('Value','Correction',100)
cv2.setTrackbarPos('Low H','Thresh Output',8)
cv2.setTrackbarPos('Low S','Thresh Output',149)
cv2.setTrackbarPos('Low V','Thresh Output',170)
cv2.setTrackbarPos('High H','Thresh Output',34)
cv2.setTrackbarPos('High S','Thresh Output',255)
cv2.setTrackbarPos('High V','Thresh Output',255)
cv2.setTrackbarPos('Kernel Size','YCrCb Median Blur',3)
cv2.setTrackbarPos('Laplacian ksize','Output',3)
cv2.setTrackbarPos('Clip Limit (Blue)','Clahe',0)
cv2.setTrackbarPos('Tile Grid Size (Blue)','Clahe',4)
cv2.setTrackbarPos('Clip Limit (Green)','Clahe',0)
cv2.setTrackbarPos('Tile Grid Size (Green)','Clahe',4)
cv2.setTrackbarPos('Clip Limit (Red)','Clahe',0)
cv2.setTrackbarPos('Tile Grid Size (Red)','Clahe',4)
cv2.setTrackbarPos('Pair','Clahe',1)
cv2.setTrackbarPos('Kernel Size','Gaussian',3)
cv2.setTrackbarPos('Standard Deviation','Gaussian',5)
cv2.setTrackbarPos('+ve Alpha','Gaussian',11)
cv2.setTrackbarPos('+ve Beta','Gaussian',2)
cv2.setTrackbarPos('-ve Alpha','Gaussian',2)
cv2.setTrackbarPos('-ve Beta','Gaussian',0)
cv2.setTrackbarPos('Alpha','Anisotropic Diffusion',1)
cv2.setTrackbarPos('Sensitivity','Anisotropic Diffusion',20)
cv2.setTrackbarPos('Iterations','Anisotropic Diffusion',2)
cv2.setTrackbarPos('Kernel','Median Blur',5)
cv2.setTrackbarPos('Kernel','Median Blur 2',3)
cv2.setTrackbarPos('FGauss','Gaussian',0)
cv2.setTrackbarPos('FGamma','Gamma Correction',1)
cv2.setTrackbarPos('FAWBal','Auto White Balance',0)
cv2.setTrackbarPos('FAni','Anisotropic Diffusion',1)
cv2.setTrackbarPos('FB','Median Blur',1)
cv2.setTrackbarPos('FG','Median Blur',1)
cv2.setTrackbarPos('FR','Median Blur',1)
cv2.setTrackbarPos('FMed','Median Blur 2',0)
cv2.setTrackbarPos('FL','Clahe LAB',0)
cv2.setTrackbarPos('FA','Clahe LAB',0)
cv2.setTrackbarPos('FB','Clahe LAB',0)
cv2.setTrackbarPos('FB','Clahe',1)
cv2.setTrackbarPos('FG','Clahe',1)
cv2.setTrackbarPos('FR','Clahe',1)
cv2.setTrackbarPos('FL','Output',0)
cv2.setTrackbarPos('FM','YCrCb Median Blur',0)
ret = True
flag=1
xdiff=0
txdiff=0
cX=0
cY=0
maxArea=0
while (1):
maxArea=0
if flag==1:
ret,img = cap.read()
fgs = cv2.getTrackbarPos('FGauss','Gaussian')
fgm = cv2.getTrackbarPos('FGamma','Gamma Correction')
fab = cv2.getTrackbarPos('FAWBal', 'Auto White Balance')
fad = cv2.getTrackbarPos('FAni','Anisotropic Diffusion')
fmb = cv2.getTrackbarPos('FB','Median Blur')
fmg = cv2.getTrackbarPos('FG','Median Blur')
fmr = cv2.getTrackbarPos('FR','Median Blur')
fmed = cv2.getTrackbarPos('FMed','Median Blur 2')
fcll = cv2.getTrackbarPos('FL','Clahe LAB')
fcla = cv2.getTrackbarPos('FA','Clahe LAB')
fclb = cv2.getTrackbarPos('FB','Clahe LAB')
fcb = cv2.getTrackbarPos('FB','Clahe')
fcg = cv2.getTrackbarPos('FG','Clahe')
fcr = cv2.getTrackbarPos('FR','Clahe')
fyl = cv2.getTrackbarPos('FL','Output')
fym = cv2.getTrackbarPos('FM','YCrCb Median Blur')
hl = cv2.getTrackbarPos('Low H','Thresh Output')
hh = cv2.getTrackbarPos('High H','Thresh Output')
sl = cv2.getTrackbarPos('Low S','Thresh Output')
sh = cv2.getTrackbarPos('High S','Thresh Output')
vl = cv2.getTrackbarPos('Low V','Thresh Output')
vh = cv2.getTrackbarPos('High V','Thresh Output')
clipLim1=(cv2.getTrackbarPos('Clip Limit (Blue)','Clahe'))
clipLim1=float(clipLim1)/1000
tgs1=cv2.getTrackbarPos('Tile Grid Size (Blue)','Clahe')
clipLim2=(cv2.getTrackbarPos('Clip Limit (Green)','Clahe'))
clipLim2=float(clipLim2)/1000
tgs2=cv2.getTrackbarPos('Tile Grid Size (Green)','Clahe')
clipLim3=(cv2.getTrackbarPos('Clip Limit (Red)','Clahe'))
clipLim3=float(clipLim3)/1000
tgs3=cv2.getTrackbarPos('Tile Grid Size (Red)','Clahe')
pgamma=cv2.getTrackbarPos('+ve Gamma','Gamma Correction')
ngamma=cv2.getTrackbarPos('-ve Gamma','Gamma Correction')
hc=cv2.getTrackbarPos('Hue','Correction')
sc=cv2.getTrackbarPos('Saturation','Correction')
vc=cv2.getTrackbarPos('Value','Correction')
pgamma=float(pgamma)/10
ngamma=float(ngamma)/10
hc=float(hc)/100
sc=float(sc)/100
vc=float(vc)/100
ks=cv2.getTrackbarPos('Kernel Size','Gaussian')
sd=(cv2.getTrackbarPos('Standard Deviation','Gaussian'))
sd=float(sd)/10
alpha=(cv2.getTrackbarPos('+ve Alpha','Gaussian'))
beta=(cv2.getTrackbarPos('+ve Beta','Gaussian'))
nalpha=(cv2.getTrackbarPos('-ve Alpha','Gaussian'))
nbeta=(cv2.getTrackbarPos('-ve Beta','Gaussian'))
alpha=float(alpha)/10
beta=float(beta)/10
nalpha=float(nalpha)/10
nbeta=float(nbeta)/10
alph=(cv2.getTrackbarPos('Alpha','Anisotropic Diffusion'))
alph=float(alph)/10
sens=cv2.getTrackbarPos('Alpha','Anisotropic Diffusion')
itern=cv2.getTrackbarPos('Alpha','Anisotropic Diffusion')
mker=cv2.getTrackbarPos('Kernel','Median Blur')
medker=cv2.getTrackbarPos('Kernel','Median Blur 2')
ymker=cv2.getTrackbarPos('Kernel Size','YCrCb Median Blur')
ylksize=cv2.getTrackbarPos('Laplacian ksize','Output')
if ks%2==0:
ks+=1
if mker%2==0:
mker+=1
if medker%2==0:
medker+=1
if ymker%2==0:
ymker+=1
if ylksize%2==0:
ylksize+=1
fg=cv2.getTrackbarPos('Pair','Clahe')
if ret:
if cv2.waitKey(2) == 27:
break
if cv2.waitKey(2) == 97:
flag = 1
if cv2.waitKey(2) == 32:
flag=0
low = np.array([hl,sl,vl])
high = np.array([hh,sh,vh])
#Gamma Correction
t1=time.time()
gc=adjust_gamma(img,pgamma-ngamma)
t2=time.time()-t1
#print(t2)
cv2.imshow('Gamma Correction',gc)
if(fgm==0):
gc=img
#Gaussian
gaussian = cv2.GaussianBlur(gc ,(ks,ks), sd)
gauss = cv2.addWeighted(img, alpha-nalpha, gaussian, beta-nbeta, 0)
cv2.imshow('Gaussian',gauss)
if(fgs==0):
gauss=gc
#Correcting HSV Values
st1=cv2.cvtColor(gauss,cv2.COLOR_BGR2HSV)
st1[:, :, 0]=st1[:, :, 0]*hc
st1[:, :, 1]=st1[:, :, 1]*sc
st1[:, :, 2]=st1[:, :, 2]*vc
st3=cv2.cvtColor(st1,cv2.COLOR_HSV2BGR)
cv2.imshow('Correction',st3)
#Auto White Balance
result1 = cv2.cvtColor(st3, cv2.COLOR_BGR2LAB)
avg_a = np.average(result1[:, :, 1])
avg_b = np.average(result1[:, :, 2])
result1[:, :, 1] = result1[:, :, 1] - ((avg_a - 128) * (result1[:, :, 0] / 255.0) * 1.5)
result1[:, :, 2] = result1[:, :, 2] - ((avg_b - 128) * (result1[:, :, 0] / 255.0) * 1.5)
result1 = cv2.cvtColor(result1, cv2.COLOR_LAB2BGR)
cv2.imshow('Auto White Balance',result1)
if (fab==0):
result1=st3
#Anisotropic Diffusion
adf = cv2.ximgproc.anisotropicDiffusion(result1, alph, sens, itern)
cv2.imshow('Anisotropic Diffusion',adf)
if(fad==0):
adf=result1
#Median Blur
b, g, r = cv2.split(adf)
b1 = cv2.medianBlur(b,mker)
if(fmb==0):
b1=b
g1 = cv2.medianBlur(g,mker)
if(fmg==0):
g1=g
r1 = cv2.medianBlur(r,mker)
if(fmr==0):
r1=r
medfil = cv2.merge((b1, g1, r1))
cv2.imshow('Median Blur',medfil)
#Clahe LAB
clahe1 = cv2.createCLAHE(clipLimit=clipLim1,tileGridSize=(tgs1,tgs1))
clahe2 = cv2.createCLAHE(clipLimit=clipLim2,tileGridSize=(tgs2,tgs2))
clahe3 = cv2.createCLAHE(clipLimit=clipLim3,tileGridSize=(tgs3,tgs3))
lab=cv2.cvtColor(medfil,cv2.COLOR_BGR2LAB)
l, a, b = cv2.split(lab)
l1 = clahe1.apply(l)
if(fcll==0):
l1=l
a1 = clahe2.apply(a)
if(fcla==0):
a1=a
b1 = clahe3.apply(b)
if(fclb==0):
b1=b
cmer=cv2.merge((l1,a1,b1))
cllab=cv2.cvtColor(cmer,cv2.COLOR_LAB2BGR)
cv2.imshow('Clahe LAB',cllab)
#Clahe BGR
if(fg==1):
cv2.setTrackbarPos('Clip Limit (Green)','Clahe',int(clipLim1*1000))
cv2.setTrackbarPos('Tile Grid Size (Green)','Clahe',tgs1)
cv2.setTrackbarPos('Clip Limit (Red)','Clahe',int(clipLim1*1000))
cv2.setTrackbarPos('Tile Grid Size (Red)','Clahe',tgs1)
clahe1 = cv2.createCLAHE(clipLimit=clipLim1,tileGridSize=(tgs1,tgs1))
clahe2 = cv2.createCLAHE(clipLimit=clipLim2,tileGridSize=(tgs2,tgs2))
clahe3 = cv2.createCLAHE(clipLimit=clipLim3,tileGridSize=(tgs3,tgs3))
b, g, r = cv2.split(cllab)
b1 = clahe1.apply(b)
if(fcb==0):
b1=b
g1 = clahe2.apply(g)
if(fcg==0):
g1=g
r1 = clahe3.apply(r)
if(fcr==0):
r1=r
cl=cv2.merge((b1,g1,r1))
cv2.imshow('Clahe',cl)
#Median Blur
medblur = cv2.medianBlur(cl,medker)
cv2.imshow('Median Blur 2',medblur)
if(fmed==0):
medblur=cl
#YCrCb Laplacian
ycrcb=cv2.cvtColor(medblur, cv2.COLOR_BGR2YCR_CB)
y,cr,cb=cv2.split(ycrcb)
dst = cv2.Laplacian( y, cv2.CV_16S, ksize=ylksize)
absDst = cv2.convertScaleAbs( dst )
out=cv2.merge((absDst,cr,cb))
output=cv2.cvtColor(out, cv2.COLOR_YCR_CB2BGR)
cv2.imshow('Output',output)
if(fyl==0):
output=medblur
#YCrCb Median
med=cv2.medianBlur(absDst,ymker)
tempOut=cv2.merge((med,cr,cb))
tempOutput=cv2.cvtColor(tempOut, cv2.COLOR_YCR_CB2BGR)
cv2.imshow('YCrCb Median Blur',tempOutput)
if(fym==0):
tempOutput=output
#Masking
hsv1= cv2.cvtColor(output, cv2.COLOR_BGR2HSV)
obj1 = cv2.inRange(hsv1, low, high)
res1 = cv2.bitwise_and(output, output, mask=obj1)
#DETECTION
gray=cv2.cvtColor(res1, cv2.COLOR_BGR2GRAY)
ret1,thresh = cv2.threshold(gray,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
(contours,hierarchy) = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
for pic, contour in enumerate(contours):
area1 = cv2.contourArea(contour)
if(area1>700):
x,y,w,h = cv2.boundingRect(contour)
if(w*1.5<h and h>120):
cv2.rectangle(res1,(x,y),(x+w,y+h),(0,165,255),2)
M = cv2.moments(contour)
cX = int(M["m10"] / M["m00"])
cY = int(M["m01"] / M["m00"])
cv2.circle(res1, (cX, cY), 7, (255, 255, 255), -1)
if(area1>maxArea):
maxArea=area1
txdiff=width2-cX
if(txdiff!=xdiff):
xdiff=txdiff
print('Pixel Difference:', xdiff)
#Display
cv2.imshow('Thresh Output',res1)
cv2.imshow('Video',img)
#VideoFileOutput.write(res1)
#VideoFileOutput2.write(img)
else:
break
cv2.destroyAllWindows()
#VideoFileOutput.release()
#VideoFileOutput2.release()
cap.release()
def get_landmarksDIST(pic):
landmarks_vectorised1 = []
#detector = dlib.get_frontal_face_detector()
frame = cv2.imread(pic)
frame = cv2.resize(frame, (300, 300))
#frame = cv2.resize(frame, (256, 256))
#image = imutils.resize(frame, width=40) # , length = 40)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
'''equ = cv2.equalizeHist(gray)
res = np.hstack((gray, equ)) # stacking images side-by-side'''
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(16, 16))
clahe_image = clahe.apply(gray)
####
clahe_image = gray
####
rects = detector(clahe_image, 1)
# loop over the face detections
#print(pic)
#print("Number of faces detected: {}".format(len(rects)))
winSize = (64, 64)
blockSize = (16, 16) #16
blockStride = (8, 8)
cellSize = (8, 8)
nbins = 9
derivAperture = 1
winSigma = 4.
histogramNormType = 0
L2HysThreshold = 2.0000000000000001e-01
gammaCorrection = 0
nlevels = 64
#hog = cv2.HOGDescriptor(winSize, blockSize, blockStride, cellSize, nbins, derivAperture, winSigma,
# histogramNormType, L2HysThreshold, gammaCorrection, nlevels)
hog = cv2.HOGDescriptor()
########
desc = LocalBinaryPatterns(250, 8)
radius = 16
# Number of points to be considered as neighbourers
no_points = 16 #8 * radius
# Uniform LBP is used
counter = 0
#########
for rect in rects:
face = 1
# extract the ROI of the *original* face, then align the face
# using facial landmarks
(x, y, w, h) = rect_to_bb(rect)
d = rect
img = clahe_image # np.float32(frame) / 255.0;
crop = img[d.top():d.bottom(), d.left():d.right()]
faceAligned = clahe_image # fa.align(frame, gray, rect)
#faceAligned = fa.align(frame, gray, rect)
#faceAligned = fa.align(frame, gray, rect)
winStride = (8, 8)
padding = (8, 8)
locations = ((10, 20), )
#detections = rects #detector(image, 1)
#d = rect
#test =0
#if test == 0: #for k, d in enumerate(detections): # For all detected face instances individually
shape = predictor(
clahe_image,
rect) # Draw Facial Landmarks with the predictor class
#landmarksPoints = np.matrix([[p.x, p.y] for p in predictor(image, detections[0]).parts()])
landmarksPoints = np.matrix(
[[p.x, p.y] for p in predictor(faceAligned, rect).parts()])
##Get centre of mass
#####hog features scickit
from skimage.feature import hog as hog1
from skimage import data, color, exposure
#crop = cv2.resize(crop, (150, 150))
'''hog_image = hog1(faceAligned, orientations=8, pixels_per_cell=(32, 32), # 32
cells_per_block=(1, 1)) #, visualize=True)'''
features, hog_image = hog1(clahe_image,
orientations=8,
pixels_per_cell=(32, 32),
cells_per_block=(1, 1),
visualise=True)
#fd, hog_image = hog(image, orientations=8, pixels_per_cell=(32, 32),
# cells_per_block=(1, 1), visualize=True)
hog_features = features
hog_images = hog_image
#h = hog.compute(faceAligned, winStride, padding, locations)
hog_image = hog_image # np.float32(hog_features) #hog_image)
h = hog_image
h = h.flatten()
image = clahe_image
'''fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(8, 4), sharex=True, sharey=True)
ax1.axis('off')
ax1.imshow(image, cmap=plt.cm.gray)
ax1.set_title('Input image')
ax1.set_adjustable('box-forced')
# Rescale histogram for better display
#hog_image_rescaled = exposure.rescale_intensity(hog_image, in_range=(0, 10))
hog_image_rescaled = hog_image
ax2.axis('off')
ax2.imshow(hog_image_rescaled, cmap=plt.cm.gray)
ax2.set_title('Histogram of Oriented Gradients')
ax1.set_adjustable('box-forced')
plt.show()'''
######get local binary pattern
#histLBP = desc.describe(crop)
# Calculate the histogram
lbp = local_binary_pattern(crop, no_points, radius,
method='uniform') #''default')
lbparray = np.array(lbp)
import scipy
histogram = scipy.stats.itemfreq(lbp)
lbparray = lbparray.flatten() #histogram.flatten()
x = itemfreq(lbp.ravel())
# Normalize the histogram
histLBP = lbparray # x[:, 1] / sum(x[:, 1])
# Append image path in X_name
#######################
xlist = []
ylist = []
allcoords = []
for i in range(1, 68): # Store X and Y coordinates in two lists
xlist.append(float(shape.part(i).x))
ylist.append(float(shape.part(i).y))
coords = [float(shape.part(i).x), float(shape.part(i).y)]
allcoords.append(coords)
###### TODO write ANIMA inspired code here
# get centre of left brows
# get centre of left eye
#distance between brows and eyes centres (add to features array)
#get centre right brows
#get centre right eyes
#distance between brows and eyes centres (add this to feature array )
#distance between centre left brows and right brows
#get centre of upper lip
#get centre of lower lips
#get distance between centres upper and lower lips
#################################
xmean = np.mean(
xlist) # Get the mean of both axes to determine centre of gravity
ymean = np.mean(ylist)
xcentral = [
(x - xmean) for x in xlist
] # get distance between each point and the central point in both axes
ycentral = [(y - ymean) for y in ylist]
if xlist[26] == xlist[
29]: # If x-coordinates of the set are the same, the angle is 0, catch to prevent 'divide by 0' error in function
anglenose = 0
else:
anglenose = int(
math.atan((ylist[26] - ylist[29]) / (xlist[26] - xlist[29])) *
180 / math.pi)
if anglenose < 0:
anglenose += 90
else:
anglenose -= 90
landmarks_vectorised = []
centreMass = ndimage.measurements.center_of_mass(np.array(allcoords))
'''p1, p2 = HeadPoseEstimation(faceAligned, landmarksPoints)
p1 = np.asarray(p1)
p2 = np.asarray(p2)'''
for x, y, w, z in zip(xcentral, ycentral, xlist, ylist):
landmarks_vectorised.append(w)
landmarks_vectorised.append(z)
meannp = np.asarray((ymean, xmean))
coornp = np.asarray((w, z))
#TODO : DISTANCE FROM CENTRE GRAVITY TO EACH POINT
#DistCentreMass = np.linalg.norm(coornp - centreMass)
DistCentreMass = np.linalg.norm(coornp - meannp)
#add points for head pose and distance for each point
'''distp1 = np.linalg.norm(coornp - p1)
distp1X = p1[0]
distp1Y = p1[1]
distp2 = np.linalg.norm(coornp - p2)
distp2X = p2[0]
distp2Y = p2[1]'''
#sqrt((x2 - x1) ** 2 + (y2 - y1) ** 2)
dist = np.linalg.norm(coornp - meannp)
anglerelative = (math.atan(
(z - ymean) / (w - xmean)) * 180 / math.pi) - anglenose
#landmarks_vectorised.append(dist)
'''landmarks_vectorised.append(distp1)
landmarks_vectorised.append(distp2)'''
'''landmarks_vectorised.append(distp1X)
landmarks_vectorised.append(distp1Y)
landmarks_vectorised.append(distp2X)
landmarks_vectorised.append(distp2Y)'''
#landmarks_vectorised.append(DistCentreMass)
#landmarks_vectorised.append(dist)
#landmarks_vectorised.append(anglerelative)
'''cv2.circle(frame, (int(coornp[0]), int(coornp[1])), 1, (255, 0, 0), -1)
p1 = (int(coornp[0]), int(coornp[1]))
#p2 = (int(centreMass[0]), int(centreMass[1]))
p2 = (int(xmean), int(ymean))
if counter < 2:
cv2.line(frame, p1, p2, (255,0, 0), 1)
counter +=1'''
###cv2.imwrite('CentreMassCK.png', frame)
if len(rects) < 1:
landmarks_vectorised2 = "error"
else:
#landmarks_vectorised2 = np.array(landmarks_vectorised)
landmarks_vectorised2 = h # np.array(h) #np.concatenate((landmarks_vectorised2, h), axis=0)
#landmarks_vectorised2 = np.array(histLBP)
#landmarks_vectorised2 = np.concatenate((landmarks_vectorised2, h), axis=0)
#landmarks_vectorised2 = landmarks_vectorised2.reshape(67,6,1)
#landmarks_vectorised1.append(landmarks_vectorised2)
#landmarks_vectorised1 = np.array(landmarks_vectorised1)
#landmarks_vectorised1 = np.array(landmarks_vectorised2)
return landmarks_vectorised2
def clache(img):
gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
clache = cv.createCLAHE(clipLimit=2, tileGridSize=(8, 8))
dst = clache.apply(gray)
cv.imshow('Clache', dst)
def CLAHE(image, clipLimit=2.0, tileGridSize=(8,8)):
clahe = cv2.createCLAHE(clipLimit=clipLimit, tileGridSize=tileGridSize)
cl_img = clahe.apply(image)
return cl_img
# Image enhancement code using CLAHE
import cv2
# Read the image
file_name = input('Enter the file name that you want to enhance: ')
# img = cv2.imread('kids-room.jpg')
img = cv2.imread(file_name)
# Preparation for applying CLAHE
clahe = cv2.createCLAHE()
# Converting the image into Grey-scale image
gray_image = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# Apply enhancement process
enh_image = clahe.apply(gray_image)
# Save it in a file
# cv2.imwrite('kids-room-enhanced.jpg', enh_image)
cv2.imwrite(file_name[:-4] + '-enhanced' + file_name[-4:], enh_image)
print('Done enhancing')
def clahe_augment2(im, tg, cl):
# applies the open CV implementation of Contrast-limited adaptive histogram equalization (CLAHE)
# see https://docs.opencv.org/3.1.0/d5/daf/tutorial_py_histogram_equalization.html
clahe_low = cv.createCLAHE(clipLimit=cl, tileGridSize=tg)
img_low = clahe_low.apply(im)
return img_low
width = gray.shape[1]
gray_counter = np.zeros(256, dtype = int) #256 from 0 to 255 frequency
for i in range(height):
for j in range(width):
gray_counter[gray[i][j]] += 1
equ_image=cv2.equalizeHist(gray)
equ_counter = np.zeros(256, dtype = int)
for i in range(height):
for j in range(width):
equ_counter[equ_image[i][j]] += 1
clahe = cv2.createCLAHE(clipLimit = 8, tileGridSize=(4,4)) #Default = clip size 40, tile grid size = 8x8
clahe_image = clahe.apply(gray)
clahe_counter = np.zeros(256, dtype = int)
for i in range(height):
for j in range(width):
clahe_counter[clahe_image[i][j]] += 1
#visualization
plt.plot(clahe_counter,'r',label="Histogram")
plt.title('Clahe-Histogram')
plt.ylabel('Intensity'),plt.xlabel('Pixel')
plt.show()
plt.plot(equ_counter,'g',label="Histogram")
plt.title('Histogram-Equalizer')
print(image_width)
print(image_height)
img = cv2.resize(img, (image_width, image_height),
interpolation=cv2.INTER_CUBIC)
img2 = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
#img2 = cv2.medianBlur(img2,3)
#img2 = cv2.bilateralFilter(img2,3,25,25)
#img2[:, :, 0] = cv2.equalizeHist(img2[:, :, 0])
#img2[:, :, 1] = cv2.equalizeHist(img2[:, :, 1])
#img2[:, :, 0] = cv2.GaussianBlur(img2[:, :, 0],(9,9),0)
#img2[:, :, 2] = cv2.equalizeHist(img2[:, :, 2])
clahe1 = cv2.createCLAHE(clipLimit=1, tileGridSize=(5, 5))
clahe2 = cv2.createCLAHE(clipLimit=1.250, tileGridSize=(25, 25))
img2[:, :, 1] = clahe1.apply(img2[:, :, 1])
img2[:, :, 2] = clahe2.apply(img2[:, :, 2])
#img2[:, :, 1] = cv2.medianBlur(img2[:, :, 1],9)
#img2[:, :, 2] = cv2.bilateralFilter(img2[:, :, 2],3,25,25)
#img2[:, :, 1] = cv2.GaussianBlur(img2[:, :, 1],(9,9),0)
img2 = cv2.cvtColor(img2, cv2.COLOR_HSV2BGR)
img3 = np.hstack((img, img2))
cv2.imshow('img3', img3)
if cv2.waitKey(0) & 0xff == 27:
cv2.destroyAllWindows()
def hist_eq(frame):
lab = cv2.cvtColor(frame, cv2.COLOR_BGR2LAB)
clahe = cv2.createCLAHE(clipLimit=4)
lab[..., 0] = clahe.apply(lab[..., 0])
frame = cv2.cvtColor(lab, cv2.COLOR_LAB2BGR)
return frame
def _clahe(img, clip_limit=40.0, tile_grid_size=(8, 8)):
clahe = cv2.createCLAHE(clip_limit, tile_grid_size)
return clahe.apply(np.array(img, dtype=np.uint8))
def process_images(self):
if (self.image_list):
# Initialize Process Counter
curr = 0
# Initialize Hash List
self.hashes = []
# Initalize Blurred Array
self.blurred = []
# Load Image Data Map
image_data = load_image_data(self.image_path)
# Error Check Image Data Map
if (image_data is None):
image_data = {}
# Calculate Hash Values
for image in self.image_list:
# Create Data Object
if (not (image in image_data)):
image_data[image] = data(lmod=util.get_lmod(image))
# Calculate Imagehash
self.hashes.append(imagehash.average_hash(Image.open(image)))
# End Imagehash Calculation----------------------------------------------------------------------------------------------------------------------------
# Store Image Name
input_image = image
# Store Recent Modification Time
curr_lmod = util.get_lmod(image)
# Calculate Blur Coefficient
if ((image_data[image].vari is None)
or (image_data[image].nmax is None)
or (image_data[image].rmsv is None)
or (image_data[image].lmod < curr_lmod)):
# Compute RMS Value
loaded_image = Image.open(image).convert('L')
image_stats = ImageStat.Stat(loaded_image)
image_rms = image_stats.rms[0]
# Determine RMS Deficiency
if (image_rms < self.rms_threshold):
# Create Cache Folder
try:
util.mkdir(
util.form_path([self.image_path, TEMP_FOLD]))
except FileExistsError:
pass
# Create Cache File
input_image = util.form_path([
util.dirname(util.absolute(image)), TEMP_FOLD,
EQ_IMAGE.format(util.basename(image))
])
# Equalize Image Histogram
image_file = cv2.imread(image, cv2.IMREAD_GRAYSCALE)
clahe = cv2.createCLAHE(clipLimit=1.125,
tileGridSize=(4, 4))
eq_image = clahe.apply(image_file)
cv2.imwrite(input_image, eq_image)
# Ignore Future Warnings
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
# Compute Laplace Matrix
loaded_image = rgb2gray(io.imread(input_image))
laplace_data = laplace(loaded_image, ksize=10)
# Store Image Data
image_data[image].vari = variance(laplace_data)
image_data[image].nmax = np.amax(laplace_data)
image_data[image].rmsv = image_rms
image_data[image].lmod = curr_lmod
# Group Blurry Images
if ((image_data[image].vari < self.var_threshold)
and (image_data[image].nmax < self.max_threshold)):
self.blurred.append(image)
# Update Prompt
print("\rProcessing Images - {}% ".format(
int(curr * 100 / len(self.image_list))),
end="")
curr += 1
# End Variance Computation---------------------------------------------------------------------------------------------------------------------------------
# Write Computed Data To Data File
with open(util.form_path([self.image_path, DATA_FILE]),
'w') as data_file:
for image in image_data:
if (image in self.image_list):
data_file.write("{},{},{},{},{}\n".format(
image, image_data[image].vari,
image_data[image].nmax, image_data[image].rmsv,
image_data[image].lmod))
# Close File
data_file.close()
# End Write Operation--------------------------------------------------------------------------------------------------------------------------------------
# Initialize Diff List
self.hash_diffs = []
# Calculate Hash Differences
for i in range(len(self.hashes) - 1):
self.hash_diffs.append(
(self.hashes[i + 1] - self.hashes[i]) * self.precision)
# End Hash Difference Computation--------------------------------------------------------------------------------------------------------------------------
# Update Prompt
print("\rProcessed All Images ")
else:
util.perror("spectra: Found no images to process")
def grab_webcamframe():
ret, frame = video_capture.read()
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
clahe_image = clahe.apply(gray)
return clahe_image
def clahe(img, clipLimit=2.0, tileGridSize=(5, 5)):
img_yuv = cv2.cvtColor(img, cv2.COLOR_RGB2LAB)
clahe = cv2.createCLAHE(clipLimit=clipLimit, tileGridSize=tileGridSize)
img_yuv[:, :, 0] = clahe.apply(img_yuv[:, :, 0])
img_output = cv2.cvtColor(img_yuv, cv2.COLOR_LAB2RGB)
return img_output
def process_img(img_path):
global avg_error, max_error, img_maxerr
# abre a imagem e extrai o tamanho dela, extensão e nome do arquivo
img = cv2.imread(img_path)
img_name = img_path.split(".")[0]
img_h, img_w, img_channels = img.shape # as imagens podem ter diferentes resoluções
# define crop e masks dependendo da resolução (usar apenas na versão Fresh Up?)
if(img_w == 1024):
crop_y = img_h - 145
crop_x = img_w - 131
size = 108
mask_img = '_mask1024.png'
imask_img = '_imask1024.png'
magic_x = 102
magic_y = 42
if(img_w == 640):
crop_y = img_h - 91
crop_x = img_w - 83
size = 68
mask_img = '_mask640.png'
imask_img = '_imask640.png'
magic_x = 64
magic_y = 26
# define alguns caminhos importantes - melhor legibilidade no código
crop_path = output_dir + img_name + '_1_crop.png'
output_imgpath = output_dir + img_name + "_"
printstep("processando %s..." % img_name)
# extrai um quadrado de 68x68 no canto inferior direito - funciona para todas as resoluções
crop_img = img[crop_y:crop_y+size, crop_x:crop_x+size]
cv2.imwrite(crop_path, crop_img)
# aplica máscara
mask_img = cv2.imread(mask_img, 0)
proc_img = cv2.bitwise_and(crop_img, crop_img, mask=mask_img)
color_value = proc_img[magic_y, magic_x] # ponto para pegar cor para pintar a mask
proc_img[numpy.where((proc_img==[0,0,0]).all(axis=2))] = color_value
cv2.imwrite(output_imgpath + "2_mask.png", proc_img)
# converte para outro espaço de cores
proc_img = cv2.cvtColor(proc_img, cv2.COLOR_BGR2Lab)
proc_img = proc_img[:,:,2] # extrai o canal
cv2.imwrite(output_imgpath + "3_colorspace.png", proc_img)
# inpainting
inmask_img = cv2.imread(imask_img, 0)
proc_img = cv2.inpaint(proc_img, inmask_img, 1, cv2.INPAINT_NS)
cv2.imwrite(output_imgpath + "4_inpaint.png", proc_img)
# normaliza com CLAHE para melhorar o resultado do inpainting
clahe = cv2.createCLAHE(clipLimit=1.0, tileGridSize=(1,1))
proc_img = clahe.apply(proc_img)
cv2.imwrite(output_imgpath + "5_clahe.png", proc_img)
# binariza a imagem usando otsu
threshold_value, proc_img = cv2.threshold(proc_img, 0, 255, cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)
cv2.imwrite(output_imgpath + "6_binary.png", proc_img)
# aplica morfologia
kernel = numpy.ones((3,3), numpy.uint8)
proc_img = cv2.morphologyEx(proc_img, cv2.MORPH_OPEN, kernel, iterations=2)
cv2.imwrite(output_imgpath + "7_morph.png", proc_img)
# detecta e desenha o contorno, apenas para fins de visualização
contours, hierarchy = cv2.findContours(proc_img, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)
contorno_img = cv2.imread(crop_path)
cv2.drawContours(contorno_img, contours, -1, (0,255,0), 1)
cv2.imwrite(output_imgpath + "8_contorno.png", contorno_img)
# desenha a elipse, apenas para fins de visualização
contours = contours[0]
ellipse = cv2.fitEllipse(contours)
cv2.ellipse(crop_img, ellipse, (0,255,0), 1)
cv2.imwrite(output_imgpath + "9_elipse.png", crop_img)
# resultado
printresult("calculando %s..." % img_name)
ang_obtido = ellipse[2]
if(ang_obtido > 90):
ang_obtido = 180 - ang_obtido
print(ang_obtido)
# coisas para o test dataset
if(glob_testmode):
ang_esperado = int(img_name.split("_")[0])
ang_esperado /= 10
error = abs(ang_esperado - ang_obtido)
printresult("ang esp: %.1f" % ang_esperado)
printresult("ang obt: %.1f" % ang_obtido)
printresult("erro: %.1f" % error)
avg_error += error
quad_imgs[int(img_name.split("_")[2])-1] += 1
quad_err[int(img_name.split("_")[2])-1] += error
if(error > max_error):
max_error = error
img_maxerr = img_name
if(glob_printstep or glob_printresult): # imprime uma newline entre cada iteração do loop se estiver debugando
print()
top_w = top_w + tw
left_w = left_w + lw
cv2.namedWindow('protus', cv2.WINDOW_NORMAL)
cv2.moveWindow('protus', top_w, left_w)
cv2.resizeWindow('protus', (int(newiw * sc), int(ih * sc)))
top_w = top_w + tw
left_w = left_w + lw
cv2.namedWindow('clahe', cv2.WINDOW_NORMAL)
cv2.moveWindow('clahe', top_w, left_w)
cv2.resizeWindow('clahe', (int(newiw * sc), int(ih * sc)))
# create a CLAHE object (Arguments are optional)
#clahe = cv2.createCLAHE(clipLimit=0.8, tileGridSize=(5,5))
clahe = cv2.createCLAHE(clipLimit=0.8, tileGridSize=(2, 2))
cl1 = clahe.apply(frame)
#image leger seuils
frame1 = np.copy(frame)
Seuil_bas = np.percentile(frame, 25)
Seuil_haut = np.percentile(frame, 99.9999)
frame1[frame1 > Seuil_haut] = Seuil_haut
#print('seuil bas', Seuil_bas)
#print('seuil haut', Seuil_haut)
fc = (frame1 - Seuil_bas) * (65000 / (Seuil_haut - Seuil_bas))
fc[fc < 0] = 0
frame_contrasted = np.array(fc, dtype='uint16')
cv2.imshow('sun', frame_contrasted)
#cv2.waitKey(0)
def transform(imagePath,show=False):
# 读取输入
image = cv2.imread(imagePath)
###### 预处理 ##########
#灰度
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
#clahe = cv2.createCLAHE(clipLimit=3.0, tileGridSize=(5,5))
#res_clahe = clahe.apply(gray.copy())
# gray = cv2.GaussianBlur(gray, (5, 5), 0)
#gray = cutHist(gray,gray[6][6] - 100, gray[6][6] - 55, 80)
if drawHist:
cv2.imshow("gray", gray)
cv2.waitKey(time)
cv2.destroyAllWindows()
drawHistoGram(gray)
return
#锐化：也叫边缘增强
kernel = np.array([[-0.7 * sharpen, -sharpen, -0.7 * sharpen], [-sharpen, 7 * sharpen , -sharpen], [-0.7 * sharpen, -sharpen, -0.7 * sharpen]], np.float32)
#np.array([[0, -1.4, 0], [-1.4, 6.2, -1.4], [0, -1.4, 0]], np.float32)
dst = cv2.filter2D(gray, -1, kernel=kernel)
bg = dst[bgx][bgy]
#二值化
ret, thresh = cv2.threshold(dst, bg, 255, cv2.THRESH_BINARY)
if debugID:
cv2.imshow("threshID", thresh)
cv2.waitKey(time)
cv2.destroyAllWindows()
#水印预处理 64 - 145(bg170, peak 164)
wp = cv2.imread(path + 'waterprint/thresh2.jpg',cv2.IMREAD_GRAYSCALE)
r, wp = cv2.threshold(wp, 127, 255, cv2.THRESH_BINARY_INV)
wp = wp // 255
if getWP:
retW, wp = cv2.threshold(wp, 178, 255, cv2.THRESH_BINARY)
cv2.imshow("wp",wp)
cv2.imwrite(path + 'waterprint/thresh2.jpg',wp)
wContour = cv2.findContours(wp, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)[1]
wContour = sorted(wContour, key=cv2.contourArea, reverse=True)[:2]
rect = cv2.minAreaRect(wContour[0])
box = cv2.boxPoints(rect)
pts = np.int0(box)
print(pts)
cv2.waitKey(time)
#clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(4,4))
#res_clahe = clahe.apply(gray.copy())
###边缘（未使用）
#edged = cv2.Canny(thresh, 120, 170)
# 展示预处理结果
if show:
print("STEP 1: 边缘检测")
result = np.hstack((dst,thresh))
cv2.imshow("Result", result)
# cv2.imshow("Edged", edged)
cv2.waitKey(time)
cv2.destroyAllWindows()
## card轮廓检测 #####################################################################################################################
#这里获取所有轮廓的坐标集合
cnts = cv2.findContours(thresh.copy(), cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)[1]
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)[:12]
screenCnt = []
# 遍历轮廓，并排序
# 计算IDcard轮廓矩形近似
cnts = rectBound(cnts,screenCnt,True)
screenCnt.clear()
#按面积和重心确定两个轮廓
findContWithArea(image,cnts,stdArea,areaOffSet,screenCnt,debugID)
### 水印轮廓 #########################################################################################################################
wpCnt = []
for sc in screenCnt:
#sc = np.array(sc)
waterCnt = []
wraped = four_point_transform(gray, sc)
#print(wraped.shape)
#wraped = np.rollaxis(wraped,1)
#print(wraped.shape)
#十字方向边缘加强
clahe = cv2.createCLAHE(clipLimit = 3.0, tileGridSize=(5,5))
res_clahe = clahe.apply(wraped.copy())
threshR = cv2.filter2D(res_clahe, -1, kernel = np.array([[0, -1.3, 0], [-1.3, 6, -1.3], [0, -1.3, 0]], np.float32))
if debugWP:
cv2.imshow("sharpen", threshR)
cv2.waitKey(time)
cv2.destroyAllWindows()
#二值化
#retR, threshR = cv2.threshold(threshR, 20 , 255, cv2.THRESH_TOZERO)
kernel_hat = np.ones((3,3),np.uint8)
alpha = getBgG(threshR)
retR, threshR = cv2.threshold(threshR, alpha - 50, 255, cv2.THRESH_BINARY)
#threshR = cv2.dilate(threshR,kernel_hat,iterations = 1)
#threshR = cv2.erode(threshR,kernel_hat,iterations = 1)
if debugWP:
cv2.imshow("binary", threshR)
cv2.waitKey(time)
cv2.destroyAllWindows()
#形态学膨胀与梯度操作
kernel_v = np.array([[0,1,0],[0,1,0],[0,1,0]],np.uint8)
kernel_h = np.array([[0,0,0],[1,1,1],[0,0,0]],np.uint8)
threshR = cv2.dilate(threshR,kernel_hat,iterations = 1)
cntsR = cv2.findContours(threshR.copy(), cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)[1]
cntsR = sorted(cntsR, key=cv2.contourArea, reverse=True)[:8]
# 计算waterprint轮廓矩形近似
cntsR = rectBound(cntsR,waterCnt)
waterCnt.clear()
for pts in cntsR:
# 计算面积及重心
m1 = cv2.moments(pts)
a1 = cv2.contourArea(pts)
if debugWP:
print(a1)
print(cv2.matchShapes(wpCntIn, pts, 1, 0.0))
copy = wraped.copy()
#cv2.drawContours(copy, [wpCntIn], 0, (0, 200, 255), 2)
#cv2.drawContours(copy, [pts], 0, (0, 255, 0), 2)
wpDebug = four_point_transform(copy, pts)
cv2.imshow("WaterPrintOutline", wpDebug)
cv2.waitKey(time)
cv2.destroyAllWindows()
#按面积确定水印轮廓
findWP(a1,m1,pts,waterCnt)
#备选
if len(waterCnt) < 1 :
threshR2 = cv2.morphologyEx(threshR, cv2.MORPH_GRADIENT, kernel_hat)
if debugWP:
cv2.imshow("erode_dilate", threshR2)
cv2.waitKey(time)
cv2.destroyAllWindows()
#threshR = cv2.dilate(threshR,kernel_h,iterations = 3)
#threshR = cv2.erode(threshR,kernel_h,iterations = 2)
cntsR = cv2.findContours(threshR2.copy(), cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)[1]
cntsR = sorted(cntsR, key=cv2.contourArea, reverse=True)[:8]
cntsR = rectBound(cntsR,waterCnt)
waterCnt.clear()
for pts in cntsR:
m1 = cv2.moments(pts)
a1 = cv2.contourArea(pts)
if debugWP:
print(a1)
print(cv2.matchShapes(wpCntIn, pts, 1, 0.0))
copy = wraped.copy()
#cv2.drawContours(copy, [wpCntIn], 0, (0, 200, 255), 2)
#cv2.drawContours(copy, [pts], 0, (0, 255, 0), 2)
wpDebug = four_point_transform(copy, pts)
cv2.imshow("WaterPrintOutline2", wpDebug)
cv2.waitKey(time)
cv2.destroyAllWindows()
findWP(a1,m1,pts,waterCnt)
#TODO:确定水印外沿
if len(waterCnt) > 0:
alpha = getBgG(wraped)
#输出预处理
#获取背景灰度与255差值（比）
bgc = (255 - alpha) / 255
#np.ones(gray.shape,np.uint8)
#曝光
wraped = cv2.add(np.array(wraped * bgc * exposure, np.uint8), wraped)
#afterProcess = cv2.addWeighted(gray, 1, gray * bgc , exposure / 255, 0)
#cv2.imshow("曝光", waraped)
#cv2.waitKey(time)
#锐化
wraped = cv2.filter2D(wraped, -1, kernel=np.array([[-0.7 * sharpen, -sharpen, -0.7 * sharpen], [-sharpen, 7.9 * sharpen , -sharpen], [-0.7 * sharpen, -sharpen, -0.7 * sharpen]], np.float32))
#cv2.imshow("锐化", wraped)
#cv2.waitKey(time)
#r, threshWP = cv2.threshold(wraped,alpha - 1, 255, cv2.THRESH_BINARY)
if debugWP:
cv2.imshow("wraped", wraped)
cv2.waitKey(time)
cv2.destroyAllWindows()
alpha = getBgG(wraped)
###从里到外四个点，再从外到里四个点，mask一下
rect = order_points(waterCnt[0])
(tl, tr, br, bl) = rect
(TLO, TRO, BRO, BLO) = rect
(TLI, TRI, BRI, BLI) = rect
for i in range(-5, 5):
TLO = tl + np.array([ i, i])
#外灰度减内灰度
deltaTL = int(wraped[int(TLO[1])][int(TLO[0])]) - int(wraped[int(TLO[1]) + 1][int(TLO[0]) + 1])
if deltaTL > wpGrad * alpha or deltaTL < - wpTxDt * alpha and int(wraped[int(TLO[1])][int(TLO[0])]) > wpBgRatio * alpha:
TLO = TLO + np.array([ -1, -1])
break
elif i == 4:
TLO = tl + np.array([- 3, - 3])
for i in range(-5, 5):
TRO = tr + np.array([ -i, i])
deltaTR = int(wraped[int(TRO[1])][int(TRO[0])]) - int(wraped[int(TRO[1]) - 1][int(TRO[0]) + 1])
if deltaTR > wpGrad * alpha or deltaTR < - wpTxDt * alpha and int(wraped[int(TRO[1])][int(TRO[0])]) > wpBgRatio * alpha:
TRO = TRO + np.array([ 1, -1])
break
elif i == 4:
TRO = tr + np.array([ 3, - 3])
for i in range(-5, 5):
BRO = br + np.array([ -i, -i])
deltaBR = int(wraped[int(BRO[1])][int(BRO[0])]) - int(wraped[int(BRO[1]) - 1][int(BRO[0]) - 1])
if deltaBR > wpGrad * alpha or deltaBR < - wpTxDt * alpha and int(wraped[int(BRO[1])][int(BRO[0]) ]) > wpBgRatio * alpha:
BRO = BRO + np.array([ 1, 1])
break
elif i == 4:
BRO = br + np.array([ 3, 3])
for i in range(-5, 5):
BLO = bl + np.array([ i, - i])
deltaBL = int(wraped[int(BLO[1])][int(BLO[0])]) - int(wraped[int(BLO[1]) + 1][int(BLO[0]) - 1])
if deltaBL > wpGrad * alpha or deltaBL < - wpTxDt * alpha and int(wraped[int(BLO[1])][int(BLO[0]) ]) > wpBgRatio * alpha:
BLO = BLO + np.array([ -1, 1])
break
elif i == 4:
BLO = bl + np.array([ -3, 3])
for i in range(-5, 5):
TLI = tl + np.array([ -i, -i])
#内灰度减外灰度
deltaTL = int(wraped[int(TLI[1])][int(TLI[0])]) - int(wraped[int(TLI[1]) - 1][int(TLI[0]) - 1])
if deltaTL > wpGrad * alpha or deltaTL < -wpTxDt * alpha and int(wraped[int(TLI[1])][int(TLI[0])]) > wpBgRatio * alpha:
#TLI = TLI + np.array([ 1, 1])
break
elif i == 4:
TLI = tl + np.array([1, 1])
for i in range(-5, 5):
TRI = tr + np.array([ i, -i])
deltaTR = int(wraped[int(TRI[1])][int(TRI[0])]) - int(wraped[int(TRI[1]) + 1][int(TRI[0]) - 1])
if deltaTR > wpGrad * alpha or deltaTR < -wpTxDt * alpha and int(wraped[int(TRI[1])][int(TRI[0])]) > wpBgRatio * alpha:
#TRI = TRI + np.array([ -1, 1])
break
elif i == 4:
TRI = tr + np.array([-1, 1])
for i in range(-5, 5):
BRI = br + np.array([ i, i])
#内灰度减外灰度
deltaBR = int(wraped[int(BRI[1])][int(BRI[0])]) - int(wraped[int(BRI[1]) + 1][int(BRI[0]) + 1])
if deltaBR > wpGrad * alpha or deltaBR < -wpTxDt * alpha and int(wraped[int(BRI[1])][int(BRI[0])]) > wpBgRatio * alpha:
#BRI = BRI + np.array([ -1, -1])
break
elif i == 4:
BRI = br + np.array([-1, -1])
for i in range(-5, 5):
BLI = bl + np.array([ -i, i])
#内灰度减外灰度
deltaBL = int(wraped[int(BLI[1])][int(BLI[0])]) - int(wraped[int(BLI[1]) - 1][int(BLI[0]) + 1])
if deltaBL > wpGrad * alpha or deltaBL < -wpTxDt * alpha and int(wraped[int(BLI[1])][int(BLI[0])]) > wpBgRatio * alpha:
#BLI = BLI + np.array([ 1, -1])
break
elif i == 4:
BLI = bl + np.array([1, -1])
#if abs(int(wraped[int(BLO[1])][int(BLO[0])]) - int(wraped[int(bl[1])][int(bl[0])])) < 0.4 * wraped[int(bl[1])][int(bl[0])]:
cntO = np.array([TLO, TRO, BRO, BLO],np.int32)
cntI = np.array([TLI, TRI, BRI, BLI],np.int32)
mask = np.zeros(wraped.shape,np.uint8)
maskIn = cv2.bitwise_not(cv2.fillConvexPoly(mask.copy(), cntI, 255))
maskOut = cv2.fillConvexPoly(mask.copy(), cntO, 255)
cv2.bitwise_and(maskIn,maskOut.copy(),mask)
cv2.imshow("mask", cv2.bitwise_and(wraped,wraped,mask = mask))
cv2.waitKey(time)
cv2.destroyAllWindows()
#cv2.imshow("maskOut", maskOut)
#cv2.waitKey(time)
#cv2.destroyAllWindows()
if debugWP:
print("tl " + str(wraped[int(TLO[1])][int(TLO[0])]) \
+ " tr " + str(wraped[int(TRO[1])][int(TRO[0])]) \
+ " br " + str(wraped[int(BRO[1])][int(BRO[0])]) \
+ " bl " + str(wraped[int(BLO[1])][int(BLO[0])]))
copy = wraped.copy()
#wpDebug = cv2.circle(copy,(int(TLO[0]),int(TLO[1])),1, color = (0,0,0))
#wpDebug = cv2.circle(wpDebug,(int(TRO[0]),int(TRO[1])),1, color = (0,0,0))
#wpDebug = cv2.circle(wpDebug,(int(BRO[0]),int(BRO[1])),1, color = (0,0,0))
#wpDebug = cv2.circle(wpDebug,(int(BLO[0]),int(BLO[1])),1, color = (0,0,0))
#wpDebug = cv2.circle(wpDebug,(int(TLI[0]),int(TLI[1])),1, color = (0,0,0))
#wpDebug = cv2.circle(wpDebug,(int(TRI[0]),int(TRI[1])),1, color = (0,0,0))
#wpDebug = cv2.circle(wpDebug,(int(BRI[0]),int(BRI[1])),1, color = (0,0,0))
#wpDebug = cv2.circle(wpDebug,(int(BLI[0]),int(BLI[1])),1, color = (0,0,0))
#cv2.imshow("WaterPrintOutline", wpDebug)
#cv2.waitKey(time)
#cv2.destroyAllWindows()
hst = cv2.calcHist([wraped],[0],mask,[256],[0,256])[:int(0.4 * alpha)]
indexes, _ = scipy.signal.find_peaks(np.array(hst).ravel(), height=20, distance=28)
if len(indexes) == 1:
indexes2, _ = scipy.signal.find_peaks(np.array(hst[:80]).ravel(), height=6, distance=10)
indexes = np.concatenate((indexes, indexes2))
print(indexes)
Cut = cutHist(wraped,int(max(indexes) * 0.5 + min(indexes) * 0.5),int(max(indexes) * 2.7), alpha,1,maskOut)
cv2.imshow("Cut", Cut)
cv2.waitKey(time)
cv2.destroyAllWindows()
plt.plot(hst,'r')
plt.xlim([0,int(0.4 * alpha)])
plt.show()
wpCnt.extend(waterCnt)
#################### 异常 ################################################################################################################
if len(screenCnt) < 2 :
raise Exception("WARNING: IDcard PROCESS FAILED")
if len(wpCnt) < 1 :
print("WARNING: WaterPrint PROCESS FAILED: " + imagePath)
#raise Exception("WARNING: WaterPrint PROCESS FAILED")
##=======================================================================================================================================##
# 展示结果
if show:
print("STEP 2: 获取轮廓")
for sc in screenCnt:
copy = image.copy()
cv2.drawContours(copy, [sc], 0, (0, 255, 0), 2)
cv2.imshow("Outline", copy)
cv2.waitKey(time)
cv2.destroyAllWindows()
#TODO:去水印 : 1.身份证拉正。 2.找水印外围框 3. 水印mask直方图（均衡化） 4.两个峰取阈值 5.原图拉到背景色####################################
#copy = gray.copy()
#icopy = image.copy()
#if len(wpCnt) > 0 :
# for wc in wpCnt:
# waterPrint = four_point_transform(copy, wc)
#drawHistoGram(waterPrint)
#Hist = cv2.calcHist([waterPrint],[0],None,[256],[0,256])
#alpha = 0
#for i in range(1,256):
# m = max(Hist)
# if Hist[i] == m:
# alpha = i -140
# break
#wpMask = wp * alpha
#wpWrap = four_point_transform(wpMask, wc, 1, 1000, 1000)
#backGround = np.ones((1000,1000),np.uint8) * alpha
#outPut = cv2.add(copy,wpWrap)
if showprint:
#print(alpha)
cv2.imshow("WaterPrint", waterPrint)
cv2.waitKey(time)
cv2.destroyAllWindows()
#wpres = four_point_transform(icopy, waterCnt[1])
#cv2.imwrite(path + '/waterprint/waterprint.jpg', wpres)
############## 透视变换 ###########################################################################################
screenCnt = np.array(screenCnt)
warped1 = four_point_transform(afterProcess, screenCnt[0])
warped2 = four_point_transform(afterProcess, screenCnt[1])
# warped = resize(warped, height=600, width=600)
#warped = cv2.resize(warped, (600, 600))
if showID:
cv2.imshow("Scanned1", warped1)
cv2.waitKey(time)
cv2.imshow("Scanned2", warped2)
cv2.waitKey(time)
cv2.destroyAllWindows()
#image = resize(orig, height=500)
return (warped1,warped2)
def createCLAHE(frame):
frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(20, 20))
res = clahe.apply(frame)
return res
import os
import cv2
from matplotlib import pyplot as plt
imgPath = '/home/dell/downloads/Skype/blur1.jpeg'
imgStatus = os.path.isfile(imgPath)
print(imgStatus)
if imgStatus == True:
img = cv2.imread(imgPath, 0)
h, w = img.shape
print(h, w)
equalImg = cv2.equalizeHist(img) # Histogram Equalization (HE)
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(4, 4))
claheImg = clahe.apply(
img) # Contrast Limited Adaptive Histogram Equalization (CLAHE)
plt.figure(figsize=(20, 20))
plt.subplot(3, 2, 1)
plt.axis('off')
plt.imshow(img, cmap='gray')
plt.title('Snow Dog')
plt.subplot(3, 2, 2)
plt.axis('off')
plt.hist(img)
plt.title('Histogram of Snow Dog')
def __init__(self, clipLimit=2, tileSize=(8, 8)):
self.trans = cv2.createCLAHE(clipLimit=clipLimit,
tileGridSize=tileSize)
def MA(base_path, image_id):
eye = cv2.imread(base_path + image_id)
image = cv2.cvtColor(eye, cv2.COLOR_BGR2RGB)
median = cv2.medianBlur(image, 1)
green_image = median.copy() # Make a copy
green_image[:, :, 0] = 0
green_image[:, :, 2] = 0
gPixels = np.array(green_image)
# -----Converting image to LAB Color model-----------------------------------
lab = cv2.cvtColor(gPixels, cv2.COLOR_BGR2LAB)
# -----Splitting the LAB image to different channels-------------------------
l, a, b = cv2.split(lab)
# -----Applying CLAHE to L-channel-------------------------------------------
clahe = cv2.createCLAHE(clipLimit=3.0, tileGridSize=(8, 8))
cl = clahe.apply(l)
# -----Merge the CLAHE enhanced L-channel with the a and b channel-----------
limg = cv2.merge((cl, a, b))
# -----Converting image from LAB Color model to RGB model--------------------
final = cv2.cvtColor(limg, cv2.COLOR_LAB2BGR)
#############################################################################
edges = cv2.Canny(final, 70, 35)
final_gray = cv2.cvtColor(final, cv2.COLOR_BGR2GRAY)
edge_test = final_gray + edges
eye_final = edge_test
# Perform closing to find individual objects
eye_final = cv2.dilate(eye_final, (3, 3),
iterations=2) # eye_final.dilate(2)
eye_final = cv2.erode(eye_final, (3, 3), iterations=1)
eye_final = cv2.dilate(eye_final, (3, 3), iterations=4)
eye_final = cv2.erode(eye_final, (3, 3), iterations=3)
# Setup SimpleBlobDetector parameters.
params = cv2.SimpleBlobDetector_Params()
# Set Area filtering parameters
params.filterByArea = True
params.minArea = 10 # lesser the area more the number of keypoints generated
# Detect blobs.
detector = cv2.SimpleBlobDetector_create(params)
big_blobs = detector.detect(eye_final)
# create a blank image to mask
blank = np.zeros(eye.shape, np.uint8)
blobs = cv2.drawKeypoints(image, big_blobs, blank, (255, 255, 255),
cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)
blobs_gray = cv2.cvtColor(blobs, cv2.COLOR_RGB2GRAY)
eye_final = eye_final - blobs_gray
eye_final = cv2.erode(eye_final, (3, 3), iterations=1)
print(eye_final[100, 200])
print(eye_final[200, 300])
print(eye_final[299, 300])
# Setup SimpleBlobDetector parameters.
params = cv2.SimpleBlobDetector_Params()
# Change thresholds
params.minThreshold = 1
params.maxThreshold = 500
# Detect blobs.
detector = cv2.SimpleBlobDetector_create(params)
# Detect blobs.
small_blobs = detector.detect(eye_final)
#print(small_blobs)
if small_blobs:
small_blobs_detected = cv2.drawKeypoints(
image, big_blobs, blank, (255, 255, 255),
cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)
eye_final_small = eye_final - cv2.cvtColor(small_blobs_detected,
cv2.COLOR_RGB2GRAY)
# print("secondtime")
print(eye_final[100, 200])
print(eye_final[200, 300])
print(eye_final[299, 300])
return eye_final_small, blobs
return 'np'
import numpy as np
import cv2
img = cv2.imread('test2.jpg', 0)
# create a CLAHE object (Arguments are optional).
clahe = cv2.createCLAHE(clipLimit=4.0, tileGridSize=(8, 8))
cl1 = clahe.apply(img)
cv2.imwrite('eq_test2_adaptive_4.jpg', cl1)
def TrackWand():
global rval, old_frame, old_gray, p0, mask, color, ig, img, frame
try:
color = (0, 0, 255)
rval, old_frame = cam.read()
cv2.flip(old_frame, 1, old_frame)
old_gray = cv2.cvtColor(old_frame, cv2.COLOR_BGR2GRAY)
equalizeHist(old_gray)
old_gray = GaussianBlur(old_gray, (9, 9), 1.5)
dilate_kernel = np.ones(dilation_params, np.uint8)
old_gray = cv2.dilate(old_gray, dilate_kernel, iterations=1)
clahe = cv2.createCLAHE(clipLimit=3.0, tileGridSize=(8, 8))
old_gray = clahe.apply(old_gray)
# Take first frame and find circles in it
p0 = cv2.HoughCircles(old_gray,
cv2.HOUGH_GRADIENT,
3,
50,
param1=240,
param2=8,
minRadius=4,
maxRadius=15)
if p0 is not None:
p0.shape = (p0.shape[1], 1, p0.shape[2])
p0 = p0[:, :, 0:2]
mask = np.zeros_like(old_frame)
except:
print("No points found")
# Create a mask image for drawing purposes
while True:
try:
rval, frame = cam.read()
cv2.flip(frame, 1, frame)
if p0 is not None:
frame_gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
equalizeHist(frame_gray)
frame_gray = GaussianBlur(frame_gray, (9, 9), 1.5)
dilate_kernel = np.ones(dilation_params, np.uint8)
frame_gray = cv2.dilate(frame_gray,
dilate_kernel,
iterations=1)
frame_clahe = cv2.createCLAHE(clipLimit=3.0,
tileGridSize=(8, 8))
frame_gray = frame_clahe.apply(frame_gray)
# calculate optical flow
p1, st, err = cv2.calcOpticalFlowPyrLK(old_gray, frame_gray,
p0, None, **lk_params)
# Select good points
good_new = p1[st == 1]
good_old = p0[st == 1]
# draw the tracks
for i, (new, old) in enumerate(zip(good_new, good_old)):
a, b = new.ravel()
c, d = old.ravel()
# only try to detect gesture on highly-rated points (below 10)
if (i < 15):
IsGesture(a, b, c, d, i)
dist = math.hypot(a - c, b - d)
if (dist < movment_threshold):
cv2.line(mask, (a, b), (c, d), (0, 255, 0), 2)
cv2.circle(frame, (a, b), 5, color, -1)
cv2.putText(frame, str(i), (a, b),
cv2.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255))
img = cv2.add(frame, mask)
cv2.putText(img, "Press ESC to close.", (5, 25),
cv2.FONT_HERSHEY_SIMPLEX, 1.0, (255, 255, 255))
cv2.imshow("Raspberry Potter", frame)
# get next frame
rval, frame = cam.read()
# Now update the previous frame and previous points
old_gray = frame_gray.copy()
p0 = good_new.reshape(-1, 1, 2)
except IndexError:
print("Index error - Tracking")
except:
e = sys.exc_info()[0]
print("Tracking Error: %s" % e)
key = cv2.waitKey(20)
if key in [27, ord('Q'), ord('q')]: # exit on ESC
cv2.destroyAllWindows()
cam.release()
break