def read_file(self, filename):
im = Image.open(filename)
r, g, b = im.split()
r_arr = plimg.pil_to_array(r)
g_arr = plimg.pil_to_array(g)
b_arr = plimg.pil_to_array(b)
r_arr1 = r_arr.reshape(1024)
g_arr1 = g_arr.reshape(1024)
b_arr1 = b_arr.reshape(1024)
arr = np.concatenate((r_arr1, g_arr1, b_arr1))
return arr
def readFile(self, filename, file):
im = Image.open(filename)
r, g, b = im.split()
r_arr = plimg.pil_to_array(r)
g_arr = plimg.pil_to_array(g)
b_arr = plimg.pil_to_array(b)
arr = np.concatenate((r_arr, g_arr, b_arr))
label = []
for i in file:
label.append(i[0])
return arr,label
def read_file(filename):
im = Image.open(filename) # 打开一个图像
# 将图像的RGB分离
r, g, b = im.split()
# 将PILLOW图像转成数组
r_arr = plimg.pil_to_array(r)
g_arr = plimg.pil_to_array(g)
b_arr = plimg.pil_to_array(b)
# 将32*32二位数组转成1024的一维数组
r_arr1 = r_arr.reshape(1024)
g_arr1 = g_arr.reshape(1024)
b_arr1 = b_arr.reshape(1024)
# 3个一维数组合并成一个一维数组,大小为3072
arr = np.concatenate((r_arr1, g_arr1, b_arr1))
return arr
def create_heatmap(xs, ys, imageSize, blobSize, cmap):
blob = Image.new('RGBA', (blobSize * 2, blobSize * 2), '#000000')
blob.putalpha(0)
colour = 255 / int(math.sqrt(len(xs)))
draw = ImageDraw.Draw(blob)
draw.ellipse((blobSize / 2, blobSize / 2, blobSize * 1.5, blobSize * 1.5),
fill=(colour, colour, colour))
blob = blob.filter(ImageFilter.GaussianBlur(radius=blobSize / 2))
heat = Image.new('RGBA', (imageSize, imageSize), '#000000')
heat.putalpha(0)
xScale = float(imageSize - 1) / (max(xs) - min(xs))
yScale = float(imageSize - 1) / (min(ys) - max(ys))
xOff = min(xs)
yOff = max(ys)
for i in range(len(xs)):
xPos = int((xs[i] - xOff) * xScale)
yPos = int((ys[i] - yOff) * yScale)
blobLoc = Image.new('RGBA', (imageSize, imageSize), '#000000')
blobLoc.putalpha(0)
blobLoc.paste(blob, (xPos - blobSize, yPos - blobSize), blob)
heat = ImageChops.add(heat, blobLoc)
norm = Normalize(vmin=min(min(heat.getdata())),
vmax=max(max(heat.getdata())))
sm = ScalarMappable(norm, cmap)
heatArray = pil_to_array(heat)
rgba = sm.to_rgba(heatArray[:, :, 0], bytes=True)
rgba[:, :, 3] = heatArray[:, :, 3]
coloured = Image.fromarray(rgba, 'RGBA')
return coloured
def _matplotlib_pil_bug_present():
"""
Determine whether PIL images should be pre-flipped due to a bug in Matplotlib.
Prior to Matplotlib 1.2.0, RGB images provided as PIL objects were
oriented wrongly. This function tests whether the bug is present.
"""
from matplotlib.image import pil_to_array
try:
from PIL import Image
except:
import Image
from astropy import log
array1 = np.array([[1, 2], [3, 4]], dtype=np.uint8)
image = Image.fromarray(array1)
array2 = pil_to_array(image)
if np.all(array1 == array2):
log.debug("PIL Image flipping bug not present in Matplotlib")
return False
elif np.all(array1 == array2[::-1, :]):
log.debug("PIL Image flipping bug detected in Matplotlib")
return True
else:
log.warning(
"Could not properly determine Matplotlib behavior for RGB images - image may be flipped incorrectly"
)
return False
def _gaussFilter(self):
kernelSize = 7
imgMatrix = self.selectedItem.image
size = imgMatrix.shape
gaussianMatrix = MImage.pil_to_array(Image.new('L', imgMatrix.shape,
0))
matrix = gkern(kernelSize)
for i in range(size[1] - (kernelSize - 1)):
for j in range(size[0] - (kernelSize - 1)):
iS = i + (kernelSize - 1) / 2
jS = j + (kernelSize - 1) / 2
#print t[i:(i + kernelSize), j:(j + kernelSize)]
gaussianMatrix[iS, jS] = np.sum(
np.multiply(
imgMatrix[i:(i + kernelSize), j:(j + kernelSize)],
matrix))
#s[iS, jS] = np.sum(t[i:(i + kernelSize), j:(j + kernelSize)],)/kernelSize**2
item = QtGui.QStandardItem("new")
item.imageFileName = None
item.image = gaussianMatrix
Image.fromarray(gaussianMatrix).save("bliblop", "BMP")
item.pixmap = QtGui.QPixmap("bliblop")
self.model.appendRow(item)
self.ui.imageListView.setModel(self.model)
def test_load_normalization():
"""Test that we're normalizing imported images correctly."""
# High res map
map = starry.Map(30)
# Render the image with starry
map.load("jupiter")
img1 = map.render(projection="rect", res=360)
img1 = img1.flatten()
# Get the image directly
earth = os.path.join(os.path.dirname(starry.__file__), "img",
"jupiter.jpg")
grayscale_pil_image = Image.open(earth).convert("L")
img2 = pil_to_array(grayscale_pil_image)
img2 = np.array(img2, dtype=float)
img2 /= 255.0 # to [0, 1] range
img2 = np.flipud(img2) # align
img2 = img2[::2, ::4] # downsample
img2 = img2.flatten()
# Compute stats
mu1 = np.mean(img1)
sig1 = np.std(img1)
mu2 = np.mean(img2)
sig2 = np.std(img2)
print(sig2 - sig1)
# Assert within a couple percent
assert np.abs(mu1 - mu2) < 0.01
assert np.abs(sig1 - sig2) < 0.02
def _openFile(self):
fileName = QtGui.QFileDialog.getOpenFileName(self, "OpenImage", "src",
"Bitmaps (*.bmp)")
if fileName == "":
h.warn("No file selected.")
return
h.log("Loaded!")
#self.pixmapItem = QtGui.QPixmap(fileName)
#item = QtGui.QGraphicsPixmapItem(self.pixmapItem)
#self.scene.addItem(item)
item = QtGui.QStandardItem(os.path.basename(str(fileName)))
item.imageFileName = str(fileName)
orgImg = Image.open(str(fileName), mode='r').convert()
item.image = MImage.pil_to_array(orgImg.convert('L'))
item.pixmap = QtGui.QPixmap(fileName)
#item.setCheckable(True)
self.model.appendRow(item)
self.ui.imageListView.setModel(self.model)
self._checkButtons()
def create_heatmap(xs, ys, imageSize, blobSize, cmap):
blob = Image.new('RGBA', (blobSize * 2, blobSize * 2), '#000000')
blob.putalpha(0)
colour = 255 / int(math.sqrt(len(xs)))
draw = ImageDraw.Draw(blob)
draw.ellipse((blobSize / 2, blobSize / 2, blobSize * 1.5, blobSize * 1.5),
fill=(colour, colour, colour))
blob = blob.filter(ImageFilter.GaussianBlur(radius=blobSize / 2))
heat = Image.new('RGBA', (imageSize, imageSize), '#000000')
heat.putalpha(0)
xScale = float(imageSize - 1) / (max(xs) - min(xs))
yScale = float(imageSize - 1) / (min(ys) - max(ys))
xOff = min(xs)
yOff = max(ys)
for i in range(len(xs)):
xPos = int((xs[i] - xOff) * xScale)
yPos = int((ys[i] - yOff) * yScale)
blobLoc = Image.new('RGBA', (imageSize, imageSize), '#000000')
blobLoc.putalpha(0)
blobLoc.paste(blob, (xPos - blobSize, yPos - blobSize), blob)
heat = ImageChops.add(heat, blobLoc)
norm = Normalize(vmin=min(min(heat.getdata())),
vmax=max(max(heat.getdata())))
sm = ScalarMappable(norm, cmap)
heatArray = pil_to_array(heat)
rgba = sm.to_rgba(heatArray[:, :, 0], bytes=True)
rgba[:, :, 3] = heatArray[:, :, 3]
coloured = Image.fromarray(rgba, 'RGBA')
return coloured
def bg_compensate(img, sigma, splinepoints, scale):
"""Reads file, subtracts background. Returns [compensated image, background]."""
from PIL import Image
# from pylab import ceil,imshow,show
import matplotlib.pyplot as plt
import numpy #,pylab
from matplotlib.image import pil_to_array
from filter import canny
from matplotlib import cm
import cProfile
img = Image.open(img)
if img.mode=='I;16':
# 16-bit image
# deal with the endianness explicitly... I'm not sure
# why PIL doesn't get this right.
imgdata = np.fromstring(img.tostring(),np.uint8)
imgdata.shape=(int(imgdata.shape[0]/2),2)
imgdata = imgdata.astype(np.uint16)
hi,lo = (0,1) if img.tag.prefix == 'MM' else (1,0)
imgdata = imgdata[:,hi]*256 + imgdata[:,lo]
img_size = list(img.size)
img_size.reverse()
new_img = imgdata.reshape(img_size)
# The magic # for maximum sample value is 281
if img.tag.has_key(281):
img = new_img.astype(np.float32) / img.tag[281][0]
elif np.max(new_img) < 4096:
img = new_img.astype(np.float32) / 4095.
else:
img = new_img.astype(np.float32) / 65535.
else:
img = pil_to_array(img)
plt.subplot(1,3,1).imshow(img, cmap=cm.Greys_r)
plt.show()
if len(img.shape)>2:
raise ValueError('Image must be grayscale')
## Create mask that will fix problem when image has black areas outside of well
edges = canny(img, np.ones(img.shape, bool), 2, .1, .3)
ci = np.cumsum(edges, 0)
cj = np.cumsum(edges, 1)
i,j = np.mgrid[0:img.shape[0], 0:img.shape[1]]
mask = ci > 0
mask = mask & (cj > 0)
mask[1:,:] &= (ci[0:-1,:] < ci[-1,j[0:-1,:]])
mask[:,1:] &= (cj[:,0:-1] < cj[i[:,0:-1],-1])
import time
t0 = time.clock()
bg = backgr(img, mask, MODE_AUTO, sigma, splinepoints=splinepoints, scale=scale)
print(("Executed in %f sec" % (time.clock() - t0)))
bg[~mask] = img[~mask]
plt.subplot(1,3,2).imshow(img - bg, cmap=cm.Greys_r)
plt.subplot(1,3,3).imshow(bg, cmap=cm.Greys_r)
plt.show()
def _openFile(self):
fileName = QtGui.QFileDialog.getOpenFileName(self, "OpenImage", "src", "Bitmaps (*.bmp)")
if fileName == "":
h.warn("No file selected.")
return
h.log("Loaded!")
#self.pixmapItem = QtGui.QPixmap(fileName)
#item = QtGui.QGraphicsPixmapItem(self.pixmapItem)
#self.scene.addItem(item)
item = QtGui.QStandardItem(os.path.basename(str(fileName)))
item.imageFileName = str(fileName)
orgImg = Image.open(str(fileName), mode='r').convert()
item.image = MImage.pil_to_array(orgImg.convert('L'))
item.pixmap = QtGui.QPixmap(fileName)
#item.setCheckable(True)
self.model.appendRow(item)
self.ui.imageListView.setModel(self.model)
self._checkButtons()
def _matplotlib_pil_bug_present():
"""
Determine whether PIL images should be pre-flipped due to a bug in Matplotlib.
Prior to Matplotlib 1.2.0, RGB images provided as PIL objects were
oriented wrongly. This function tests whether the bug is present.
"""
from matplotlib.image import pil_to_array
try:
from PIL import Image
except:
import Image
from astropy import log
array1 = np.array([[1, 2], [3, 4]], dtype=np.uint8)
image = Image.fromarray(array1)
array2 = pil_to_array(image)
if np.all(array1 == array2):
log.debug("PIL Image flipping bug not present in Matplotlib")
return False
elif np.all(array1 == array2[::-1, :]):
log.debug("PIL Image flipping bug detected in Matplotlib")
return True
else:
log.warn("Could not properly determine Matplotlib behavior for RGB images - image may be flipped incorrectly")
return False
def getPredictionOfImage(content, names, dates, model=model):
if content is not None:
# content = "abcdeghifklmnop"
img_type = str(content).split("/")[1].replace(";base64,", "")
img_encoded = str(content).replace(
str("data:image/" + str(img_type) + ";base64,"), "")
img_decoded = base64.b64decode(img_encoded)
# Save the image into a temporary directory
temp_dir_path = loadImageIntoTempDir(img=img_decoded,
img_type=img_type)
# Create a generator object based on the contents of the temporary directory
temp_data_gen, temp_image = createGenObject(temp_dir_path)
# Assign a grade to the image in the temp dir using the model
grade = assignPrediction(temp_data_gen, model)
# Generate the final plot with the predictedgrade
fname = str(temp_dir_path + "/unlabelled/img." + img_type)
# img_high_res = mpimg.imread(fname=fname, format=img_type)
img_high_res = pil_to_array(PIL.Image.open(fname))
finalPlot = assignImgGrade(img=img_high_res, grade=grade)
# Clean up and remove the temporary directory
rmtree(temp_dir_path)
return finalPlot
return None
def bg_compensate(img, sigma, splinepoints, scale):
'''Reads file, subtracts background. Returns [compensated image, background].'''
from PIL import Image
from pylab import ceil,imshow,show
import numpy,pylab
from matplotlib.image import pil_to_array
from filter import canny
import matplotlib
import cProfile
img = Image.open(img)
if img.mode=='I;16':
# 16-bit image
# deal with the endianness explicitly... I'm not sure
# why PIL doesn't get this right.
imgdata = np.fromstring(img.tostring(),np.uint8)
imgdata.shape=(int(imgdata.shape[0]/2),2)
imgdata = imgdata.astype(np.uint16)
hi,lo = (0,1) if img.tag.prefix == 'MM' else (1,0)
imgdata = imgdata[:,hi]*256 + imgdata[:,lo]
img_size = list(img.size)
img_size.reverse()
new_img = imgdata.reshape(img_size)
# The magic # for maximum sample value is 281
if img.tag.has_key(281):
img = new_img.astype(np.float32) / img.tag[281][0]
elif np.max(new_img) < 4096:
img = new_img.astype(np.float32) / 4095.
else:
img = new_img.astype(np.float32) / 65535.
else:
img = pil_to_array(img)
pylab.subplot(1,3,1).imshow(img, cmap=matplotlib.cm.Greys_r)
pylab.show()
if len(img.shape)>2:
raise ValueError('Image must be grayscale')
## Create mask that will fix problem when image has black areas outside of well
edges = canny(img, np.ones(img.shape, bool), 2, .1, .3)
ci = np.cumsum(edges, 0)
cj = np.cumsum(edges, 1)
i,j = np.mgrid[0:img.shape[0], 0:img.shape[1]]
mask = ci > 0
mask = mask & (cj > 0)
mask[1:,:] &= (ci[0:-1,:] < ci[-1,j[0:-1,:]])
mask[:,1:] &= (cj[:,0:-1] < cj[i[:,0:-1],-1])
import time
t0 = time.clock()
bg = backgr(img, mask, MODE_AUTO, sigma, splinepoints=splinepoints, scale=scale)
print ("Executed in %f sec" % (time.clock() - t0))
bg[~mask] = img[~mask]
pylab.subplot(1,3,2).imshow(img - bg, cmap=matplotlib.cm.Greys_r)
pylab.subplot(1,3,3).imshow(bg, cmap=matplotlib.cm.Greys_r)
pylab.show()
def get_img_trasi_aphrodite_score_triple(image, model):
image_array = mpimg.pil_to_array(image)
image_array = np.asarray(image_array)
image_array = image_array[:, :, 0:3]
image_array = image_array / 255
image_array = np.expand_dims(image_array, 0)
aphrodite_score = model.predict(image_array)
return image, get_trasi_score(image), aphrodite_score
def get_patch(self):
i = random.randint(0, self.data_n - 1)
patch = Image.open((os.path.join(self.data_path, self.data_directory[i])))
im_width, im_height = patch.size
window_x = np.random.randint(im_width - self.patch_size[0])
window_y = np.random.randint(im_height - self.patch_size[1])
window = patch.crop((window_x, window_y, window_x + self.patch_size[0], window_y + self.patch_size[1]))
patch_array = mpimage.pil_to_array(window)
return patch_array
def read_file(self,filename):
im = Image.open(os.path.join("E:\年龄识别\MORPH数据库\imc_Im100_100\image17",filename),'r')#打开一个图像
site=filename.find('.')
label_arr = np.array([int(filename[site-2:site])])
# 将图像的RGB分离
r, g, b = im.split()
# 将PILLOW图像转成数组
r_arr = plimg.pil_to_array(r)
g_arr = plimg.pil_to_array(g)
b_arr = plimg.pil_to_array(b)
# 将200*240二维数组转成48000的一维数组
r_arr1 = r_arr.reshape(10000)
g_arr1 = g_arr.reshape(10000)
b_arr1 = b_arr.reshape(10000)
# 3个一维数组合并成一个一维数组,大小为144000
arr = np.concatenate((r_arr1, g_arr1, b_arr1))
return arr,label_arr
def on_activated(self, action, x1, y1, x2, y2):
if x1 and y1 and x2 and y2:
if action == 'Crop':
self.dx1.append(x1)
self.dy1.append(y1)
self.dx2.append(self.imgWidth - x2)
self.dy2.append(y1)
self.dx3.append(x1)
self.dy3.append(self.imgHeight - y2)
self.dx4.append(self.imgWidth - x2)
self.dy4.append(self.imgHeight - y2)
self.stackImgArr.append(self.imgArr)
self.dx_dy['topLeft'] = [sum(self.dx1), sum(self.dy1)]
self.dx_dy['topRight'] = [self.refImgWidth - sum(self.dx2), sum(self.dy2)]
self.dx_dy['bottomLeft'] = [sum(self.dx3), self.refImgHeight - sum(self.dy3)]
self.dx_dy['bottomRight'] = [self.refImgWidth - sum(self.dx4), self.refImgHeight - sum(self.dy4)]
self.img['src'] = self.img['src'].crop((int(x1), int(y1), int(x2), int(y2)))
self.imgWidth, self.imgHeight = self.img['src'].size
self.imgArr = mpimg.pil_to_array(self.img['src'])
self.imgArr.setflags(write=True)
self.refresh_Img_plot(self.img)
self.draw()
self.set_init_coords()
elif action == 'Delete':
x1, y1, x2, y2 = map(int, [x1, y1, x2, y2])
ImgArrCopy = deepcopy(self.imgArr)
self.stackImgArr.append(ImgArrCopy)
self.imgArr[y1:y2, x1:x2] = 0
self.img['src'] = Image.fromarray(self.imgArr)
self.dx1.append(0)
self.dx2.append(0)
self.dx3.append(0)
self.dx4.append(0)
self.dy1.append(0)
self.dy2.append(0)
self.dy3.append(0)
self.dy4.append(0)
self.refresh_Img_plot(self.img)
self.draw()
self.set_init_coords()
if action == 'Revert': #TODO: revert should store only changes to image array
if self.stackImgArr:
self.imgArr = self.stackImgArr[-1]
self.img['src'] = Image.fromarray(self.imgArr)
self.stackImgArr = self.stackImgArr[:-1]
self.imgWidth, self.imgHeight = self.img['src'].size
del self.dx1[-1]
del self.dx2[-1]
del self.dx3[-1]
del self.dx4[-1]
del self.dy1[-1]
del self.dy2[-1]
del self.dy3[-1]
del self.dy4[-1]
self.profImshow()
self.draw()
def refresh_Img_plot(self, img):
self.ax.cla()
self.img = img
self.imgArr = mpimg.pil_to_array(self.img['src'])
self.imgArr.setflags(write=True)
self.profImshow()
self.ax.set_title(self.img['pltTitle'])
self.ax.callbacks.connect('xlim_changed', self.on_xlims_change)
self.ax.callbacks.connect('ylim_changed', self.on_ylims_change)
self.fig.canvas.draw_idle()
def ask_for_removal(self, image):
if image is None:
ans = input(
'Cannot display image unfortunately. Shall we try to remove it or not? y/n: '
).lower()
else:
plt.imshow(mpImage.pil_to_array(image))
plt.show(block=False)
ans = input('Should we remove this image? y/n: ').lower()
return ans == 'y'
def read_image(filepath, ratio):
im = Image.open(filepath)
(length, height) = im.size
im = im.resize((length // ratio, height // ratio))
print(im)
# im.save('cymsbsbsbsb.png')
imarr = plimg.pil_to_array(im)
imarr.flags.writeable = True
graph_normalization(imarr, length // ratio, height // ratio)
return (imarr, length // ratio, height // ratio)
def read_file(self, filename):
im = Image.open(os.path.join("image_test1", filename), 'r') #打开一个图像
#site=filename.find('.')
#label_arr = np.array([int(filename[site-2:site])])
#label_arr = np.array([int(filename[0])])
# 将图像的RGB分离
r, g, b = im.split()
# 将PILLOW图像转成数组
r_arr = plimg.pil_to_array(r)
g_arr = plimg.pil_to_array(g)
b_arr = plimg.pil_to_array(b)
# 将224*224二维数组转成50176的一维数组
r_arr1 = r_arr.reshape(50176)
g_arr1 = g_arr.reshape(50176)
b_arr1 = b_arr.reshape(50176)
# 3个一维数组合并成一个一维数组,大小为150528
arr = np.concatenate((r_arr1, g_arr1, b_arr1))
return arr
def applyGaussianFilter(imgMatrix, kernelSize = 7):
gaussianMatrix = MImage.pil_to_array(Image.new('L', imgMatrix.shape, 0))
matrix = gkern(kernelSize)
for i in range(size[1] - (kernelSize - 1)):
for j in range(size[0] - (kernelSize - 1)):
iS = i + (kernelSize - 1)/2
jS = j + (kernelSize - 1)/2
#print t[i:(i + kernelSize), j:(j + kernelSize)]
gaussianMatrix[iS, jS] = np.sum(np.multiply(imgMatrix[i:(i + kernelSize), j:(j + kernelSize)], matrix))
#s[iS, jS] = np.sum(t[i:(i + kernelSize), j:(j + kernelSize)],)/kernelSize**2
return gaussianMatrix
def read_file(self,filename):
im = Image.open(os.path.join(r"C:\Users\yy\Desktop\Stock\line\line\line_test",filename),'r')#打开一个图像
site=filename.find('.')
label_arr = np.array([int(filename[site-1:site])])
# 将图像的RGB分离
r, g, b = im.split()
# 将PILLOW图像转成数组
r_arr = plimg.pil_to_array(r)
g_arr = plimg.pil_to_array(g)
b_arr = plimg.pil_to_array(b)
# 将200*240二维数组转成48000的一维数组
r_arr1 = r_arr.reshape(224*224)
g_arr1 = g_arr.reshape(224*224)
b_arr1 = b_arr.reshape(224*224)
# 3个一维数组合并成一个一维数组,大小为144000
arr = np.concatenate((r_arr1, g_arr1, b_arr1))
return arr,label_arr
def __init__(self, name, add_noise=True, verbose=False): """ Class Image constructor Parameters ---------- name: str Path to the image file add_noise: Bool, optional, default: True Add noise to smooth histograms of highly compressed jpegs and palette based images. verbose: Bool/str, optional, default: False Verbosity level: - False: silent run - True: report textual information - Path prefix: report textual information and report graphical information in verbose+<suffix>.jpg files """ self.verbose=verbose self.name = name self.jpeg_quality = 100 self.palette_use = 1. self.add_noise = add_noise with open(name, "rb") as fh: image = pil.Image.open(fh) if image.format == 'JPEG': q0 = np.array(image.quantization[0])[zigzag_index].reshape(8, 8) if 1 in image.quantization: q1 = np.array(image.quantization[1])[zigzag_index].reshape(8, 8) else: q1 = q0.copy() S = np.vstack([(100 * q0[:3, :3] - 50)/ijg_q0[:3, :3], (100 * q1[:3, :3] - 50)/ijg_q1[:3, :3]]).max() self.jpeg_quality = ((200 - S)/2 if S < 100 else 5000/S) if image.mode == 'P': csc = np.cumsum(np.array(sorted(image.getcolors(), key=lambda x:x[0]))[:, 0]) self.palette_use = csc[csc > csc.max()/20.].shape[0]/float(len(csc)) img = pil_to_array(image) if np.max(img) <= 1: img = np.uint8(img*255) self.valid = True if len(img.shape) < 3 or img.shape[2] < 3: self.valid = False return self.img = img[:, :, :3]
def InitImage(image):
#read PIL image (1-D image) as graph
#image = Image.open('image.jpg').convert("L")
img = mpimg.pil_to_array(image)
plt.figure(figsize=(4, 4))
plt.imshow(img, cmap='gray')
G = nx.grid_2d_graph(img.shape[0], img.shape[1])
#nx.set_node_attributes(G,{u:{'intensity':v} for u,v in np.ndenumerate(img)})
for u, v, d in G.edges(data=True):
d['weight'] = 20 - abs(np.subtract(int(img[u]), int(img[v]))) + 0.5
return G
def display(display_list):
plt.figure(figsize=(8, 8))
title = ['Input Image', 'True Mask', 'Predicted Mask']
for i in range(len(display_list)):
plt.subplot(1, len(display_list), i + 1)
plt.title(title[i])
A = tf.keras.preprocessing.image.array_to_img(display_list[i])
img = np.ma.array(pil_to_array(A))
plt.imshow(img)
plt.axis('off')
plt.show()
def applyGaussianFilter(imgMatrix, kernelSize=7):
gaussianMatrix = MImage.pil_to_array(Image.new('L', imgMatrix.shape, 0))
matrix = gkern(kernelSize)
for i in range(size[1] - (kernelSize - 1)):
for j in range(size[0] - (kernelSize - 1)):
iS = i + (kernelSize - 1) / 2
jS = j + (kernelSize - 1) / 2
#print t[i:(i + kernelSize), j:(j + kernelSize)]
gaussianMatrix[iS, jS] = np.sum(
np.multiply(imgMatrix[i:(i + kernelSize), j:(j + kernelSize)],
matrix))
#s[iS, jS] = np.sum(t[i:(i + kernelSize), j:(j + kernelSize)],)/kernelSize**2
return gaussianMatrix
def read_file(self,filename):
im = Image.open(os.path.join("./unlabelled/"+filename))#打开一个图像
im = im.resize((224, 224))
# 将图像的RGB分离
r, g, b = im.split()
# 将PILLOW图像转成数组
r_arr = plimg.pil_to_array(r)
g_arr = plimg.pil_to_array(g)
b_arr = plimg.pil_to_array(b)
# 将32*32二位数组转成1024的一维数组
r_arr1 = r_arr.reshape(50176)
g_arr1 = g_arr.reshape(50176)
b_arr1 = b_arr.reshape(50176)
# 标签
# file_labels = []
file_label = [0]
if os.path.exists(os.path.join("Glaucoma+/"+filename)):
file_label = [1]
# print("Normal:",filename)
# file_labels.append(file_label)
# 3个一维数组合并成一个一维数组,大小为244860+1
arr = np.concatenate((r_arr1, g_arr1, b_arr1))
return arr,file_label
def _pil_to_array(im):
"""
Wrapper to handle mod introduced in matplotlib 1.2 that turned array upside down.
"""
rgb = pil_to_array(im)
M, m, r = matplotlib.__version__.split('.')
ver = int(M) * 100 + int(m)
if ver > 101: # matplotlib 1.2 and later
rank = len(rgb.shape)
if rank == 2:
rgb = rgb[-1::-1, :]
else:
rgb = rgb[-1::-1, :, :]
return rgb
def read_file(self, filename, classnum):
#im = Image.open(os.path.join("../Stock/up/up_train/up_%s"%classnum,filename),'r')#打开一个图像
im = Image.open(
os.path.join(
r"C:\Users\yy\Desktop\Stock\line\line\line_train\line_%s" %
classnum, filename), 'r') # 打开一个图像
site = filename.find('.')
label_arr = np.array([int(filename[site - 1:site])])
# 将图像的RGB分离
r, g, b = im.split()
# 将PILLOW图像转成数组
r_arr = plimg.pil_to_array(r)
g_arr = plimg.pil_to_array(g)
b_arr = plimg.pil_to_array(b)
# 将224*224二维数组转成50176的一维数组
r_arr1 = r_arr.reshape(224 * 224)
g_arr1 = g_arr.reshape(224 * 224)
b_arr1 = b_arr.reshape(224 * 224)
# 3个一维数组合并成一个一维数组,大小为150528
arr = np.concatenate((r_arr1, g_arr1, b_arr1))
return arr, label_arr
def __init__(self, img, width=5, height=4, dpi=100):
self.img = img
self.imgArr = mpimg.pil_to_array(self.img['src'])
self.imgArr.setflags(write=True)
self.stackImgArr = []
self.limX = ()
self.limY = ()
self.fig = plt.figure(figsize=(width, height), dpi=dpi)
self.ax = self.fig.add_subplot(111)
self.ax.set_title(self.img['pltTitle'])
self.ax.callbacks.connect('xlim_changed', self.on_xlims_change)
self.ax.callbacks.connect('ylim_changed', self.on_ylims_change)
FigureCanvas.__init__(self, self.fig)
self.set_init_coords()
self.initCropCoords(img)
self.plot()
def frames(self, count=0):
"""
Return iterator over frames.
The composition of functions in `self.fns`
list is applied to each frame. By default, this list is empty. Examples
of function "hooks" to put into `self.fns` are functions from ``scipy.ndimage``.
"""
fn = self.pipeline()
while True:
try:
self.im.seek(count)
count += 1
yield fn(mpl_img.pil_to_array(self.im))
except EOFError:
break
def save_diff(images):
tmp = images.copy()
for k in range(len(images)):
imagename = "./save/" + str(k) + ".jpg"
plt.imsave(imagename,
images[k].reshape(28, 28),
vmin=0,
vmax=255,
format="jpg",
cmap='gray')
for k in range(len(images)):
imagename = "./save/" + str(k) + ".jpg"
image = Image.open(imagename, mode='r').convert('L')
im_array = mp.pil_to_array(image)
tmp[k] = im_array
return tmp
def image2map(image, lmax=10):
"""Return a map vector corresponding to a lat-long map image."""
# If image doesn't exist, check for it in maps directory
if not os.path.exists(image):
dn = os.path.dirname
image = os.path.join(dn(os.path.abspath(__file__)), image + ".jpg")
if not os.path.exists(image):
raise ValueError("File not found: %s." % image)
# Get the image array
grayscale_pil_image = Image.open(image).convert("L")
image_array = pil_to_array(grayscale_pil_image)
image_array = np.array(image_array, dtype=float)
image_array /= np.max(image_array)
# Convert it to a map
return array2map(image_array, lmax=lmax)
def on_file_open(self, event):
dlg = wx.FileDialog(self.frame, style=wx.FD_OPEN)
if dlg.ShowModal() == wx.ID_OK:
img = pil_to_array(
PILImage.open(
os.path.join(dlg.GetDirectory(), dlg.GetFilename())))
if img.ndim == 3:
img = img[:, :, 0] + img[:, :, 1] + img[:, :, 2]
img = stretch(img.astype(float))
lt = self.frame.MenuBar.Menus[0][0].MenuItems[7].IsChecked()
if lt:
limg, d = log_transform(img)
else:
limg = img
self.frame.subplot_imshow_grayscale(0, 0, limg)
limg = limg.flatten()
menu_items = self.frame.MenuBar.Menus[0][0].MenuItems
if menu_items[2].IsChecked():
t1 = t2 = otsu(limg)
elif menu_items[3].IsChecked():
t1 = t2 = entropy(limg)
elif menu_items[4].IsChecked():
t1, t2 = otsu3slow(limg)
elif menu_items[5].IsChecked():
t1, t2 = entropy3(limg)
else:
t1, t2 = otsu3(limg)
if lt:
t1, t2 = inverse_log_transform(np.array([t1, t2]), d)
m1 = img < t1
m2 = np.logical_and(img >= t1, img < t2)
m3 = img > t2
cimg = np.zeros((m1.shape[0], m1.shape[1], 3))
cimg[:, :, 0][m1] = img[m1]
cimg[:, :, 1][m2] = img[m2]
cimg[:, :, 2][m3] = img[m3]
self.frame.subplot_imshow(1,
0,
cimg,
sharex=self.frame.subplot(0, 0),
sharey=self.frame.subplot(0, 0))
self.frame.Refresh()
wx.MessageBox("Low threshold = %f, high threshold = %f" %
(t1, t2),
parent=self.frame)
def test_load():
"""Test that we're loading images correctly."""
# Get the image directly
image = os.path.join(os.path.dirname(starry.__file__), "img", "spot.png")
grayscale_pil_image = Image.open(image).convert("L")
img1 = pil_to_array(grayscale_pil_image)
img1 = np.array(img1, dtype=float)
img1 /= 255.0 # to [0, 1] range
img1 = np.flipud(img1) # align
img1 = img1[:, ::2] # make square to match starry output
# Render the image with starry at high res
map = starry.Map(30)
map.load("spot", extent=(-180, 179, -90, 89))
img2 = map.render(projection="rect", res=img1.shape[0])
# Canny filter on both images to detect edges
img1 = np.array(canny(img1, sigma=4), dtype=int)
img2 = np.array(canny(img2, sigma=4), dtype=int)
# Compute the difference between the images
# Do this several times by translating one of
# the images by a single pixel and computing
# the *minimum*. This has the effect of allowing
# a one-pixel tolerance in the comparison
diff = np.min(
np.abs([
img2 - img1,
img2 - shift(img1, [0, 1]),
img2 - shift(img1, [0, -1]),
img2 - shift(img1, [1, 0]),
img2 - shift(img1, [-1, 0]),
]),
axis=0,
)
# Number of pixels in the image edges
pixels = np.count_nonzero(img1)
# Number of pixels that differ between the two
pixels_diff = np.count_nonzero(diff)
# There should be VERY few pixels in the difference image
assert pixels_diff / pixels < 0.01
def on_file_open(self, event):
dlg = wx.FileDialog(self.frame,style=wx.FD_OPEN)
if dlg.ShowModal() == wx.ID_OK:
img = pil_to_array(PILImage.open(os.path.join(dlg.GetDirectory(),dlg.GetFilename())))
if img.ndim == 3:
img = img[:,:,0]+img[:,:,1]+img[:,:,2]
img = stretch(img.astype(float))
lt = self.frame.MenuBar.Menus[0][0].MenuItems[7].IsChecked()
if lt:
limg, d = log_transform(img)
else:
limg = img
self.frame.subplot_imshow_grayscale(0, 0, limg)
limg = limg.flatten()
menu_items = self.frame.MenuBar.Menus[0][0].MenuItems
if menu_items[2].IsChecked():
t1 = t2 = otsu(limg)
elif menu_items[3].IsChecked():
t1 = t2 = entropy(limg)
elif menu_items[4].IsChecked():
t1, t2 = otsu3slow(limg)
elif menu_items[5].IsChecked():
t1, t2 = entropy3(limg)
else:
t1, t2 = otsu3(limg)
if lt:
t1,t2 = inverse_log_transform(np.array([t1,t2]), d)
m1 = img < t1
m2 = np.logical_and(img >= t1, img < t2)
m3 = img > t2
cimg = np.zeros((m1.shape[0],m1.shape[1],3))
cimg[:,:,0][m1]=img[m1]
cimg[:,:,1][m2]=img[m2]
cimg[:,:,2][m3]=img[m3]
self.frame.subplot_imshow(1, 0, cimg,
sharex = self.frame.subplot(0,0),
sharey = self.frame.subplot(0,0))
self.frame.Refresh()
wx.MessageBox("Low threshold = %f, high threshold = %f"%(t1,t2),
parent=self.frame)
def _gaussFilter(self):
kernelSize = 7
imgMatrix = self.selectedItem.image
size = imgMatrix.shape
gaussianMatrix = MImage.pil_to_array(Image.new('L', imgMatrix.shape, 0))
matrix = gkern(kernelSize)
for i in range(size[1] - (kernelSize - 1)):
for j in range(size[0] - (kernelSize - 1)):
iS = i + (kernelSize - 1)/2
jS = j + (kernelSize - 1)/2
#print t[i:(i + kernelSize), j:(j + kernelSize)]
gaussianMatrix[iS, jS] = np.sum(np.multiply(imgMatrix[i:(i + kernelSize), j:(j + kernelSize)], matrix))
#s[iS, jS] = np.sum(t[i:(i + kernelSize), j:(j + kernelSize)],)/kernelSize**2
item = QtGui.QStandardItem("new")
item.imageFileName = None
item.image = gaussianMatrix
Image.fromarray(gaussianMatrix).save("bliblop", "BMP")
item.pixmap = QtGui.QPixmap("bliblop")
self.model.appendRow(item)
self.ui.imageListView.setModel(self.model)
import pylab as P
from matplotlib.toolkits.basemap import Basemap
from matplotlib.numerix import ma
from matplotlib.image import pil_to_array
from PIL import Image
# shows how to warp an image from one map projection to another.
# image from http://visibleearth.nasa.gov/
# read in jpeg image to rgba array of normalized floats.
pilImage = Image.open('land_shallow_topo_2048.jpg')
rgba = pil_to_array(pilImage)
rgba = rgba.astype(P.Float32)/255. # convert to normalized floats.
# define lat/lon grid that image spans (projection='cyl').
nlons = rgba.shape[1]; nlats = rgba.shape[0]
delta = 360./float(nlons)
lons = P.arange(-180.+0.5*delta,180.,delta)
lats = P.arange(-90.+0.5*delta,90.,delta)
# create new figure
fig=P.figure()
# define cylindrical equidistant projection.
m = Basemap(projection='cyl',llcrnrlon=-180,llcrnrlat=-90,urcrnrlon=180,urcrnrlat=90,resolution='l')
# plot (unwarped) rgba image.
im = m.imshow(rgba)
# draw coastlines.
m.drawcoastlines(linewidth=0.5,color='0.5')
# draw lat/lon grid lines.
m.drawmeridians(P.arange(-180,180,60),labels=[0,0,0,1],color='0.5')
m.drawparallels(P.arange(-90,90,30),labels=[1,0,0,0],color='0.5')
print "DONE"
return image
#######################################
print "Loading and Converting Image ...",
orgImg = Image.open("src/" + orgFileName, mode = 'r').convert()
try:
orgImg = orgImg.convert().crop((0,0,size[0],size[1]))
except NameError:
size = orgImg.size
orgGreyscaleMatrix = MImage.pil_to_array(orgImg.convert('L'))
emptyGreyscaleMatrix = MImage.pil_to_array(Image.new('L', size, 0))
emptyColoredMatrix = MImage.pil_to_array(Image.new('RGB', size, (0,0,0)))
print "DONE"
# print "Compute the 2-dimensional FFT ...",
# #print "."
# m = np.fft.rfft2(orgGreyscaleMatrix)
# #print m.shape
# #print m
# #print np.amax(m)
# am = np.log(np.absolute(m)+1)
# mm = np.around(am/np.amax(am)*255, decimals=0).astype(np.uint8)
# print "DONE"
def test_02_01_compare_to_matlab(self):
path = os.path.split(__file__)[0]
mask_file = os.path.join(path, 'Channel2-01-A-01Mask.png')
mask = pil_to_array(PILImage.open(mask_file))
mask = np.flipud(mask)
texture_measurements = loadmat(os.path.join(path,'texturemeasurements.mat'), struct_as_record=True)
texture_measurements = texture_measurements['m'][0,0]
image_file = os.path.join(example_images_directory(),
'ExampleSBSImages', 'Channel1-01-A-01.tif')
image = pil_to_array(PILImage.open(image_file))
image = np.flipud(image[:,:,0])
image = image.astype(float) / 255.0
labels,count = scind.label(mask.astype(bool),np.ones((3,3),bool))
centers = scind.center_of_mass(np.ones(labels.shape), labels,
np.arange(count)+1)
centers = np.array(centers)
X = 1 # the index of the X coordinate
Y = 0 # the index of the Y coordinate
order_python = np.lexsort((centers[:,X],centers[:,Y]))
workspace, module = self.make_workspace(image, labels, convert = False)
module.scale_groups[0].scale.value = 3
my_angle = M.H_HORIZONTAL
module.scale_groups[0].angles.value = my_angle
module.run(workspace)
m = workspace.measurements
tm_center_x = texture_measurements['Location_Center_X'][0,0][:,0]
tm_center_y = texture_measurements['Location_Center_Y'][0,0][:,0]
order_matlab = np.lexsort((tm_center_x,tm_center_y))
for measurement in M.F_HARALICK:
mname = '%s_%s_%s_%d'%(M.TEXTURE, measurement, INPUT_IMAGE_NAME, 3)
pymname = mname + "_" + M.H_TO_A[my_angle]
pytm = m.get_current_measurement(INPUT_OBJECTS_NAME, pymname)
tm = texture_measurements[mname][0,0][:,]
error_count = 0
for i in range(count):
matlab_val = tm[order_matlab[i]]
python_val = pytm[order_python[i]]
self.assertAlmostEqual(tm[order_matlab[i]],
pytm[order_python[i]],7,
"Measurement = %s, Loc=(%.2f,%.2f), Matlab=%f, Python=%f"%
(mname, tm_center_x[order_matlab[i]],
tm_center_y[order_matlab[i]],
tm[order_matlab[i]],
pytm[order_python[i]]))
image_measurements =\
(('Texture_AngularSecondMoment_Cytoplasm_3', 0.5412),
('Texture_Contrast_Cytoplasm_3',0.1505),
('Texture_Correlation_Cytoplasm_3', 0.7740),
('Texture_Variance_Cytoplasm_3', 0.3330),
('Texture_InverseDifferenceMoment_Cytoplasm_3',0.9321),
('Texture_SumAverage_Cytoplasm_3',2.5684),
('Texture_SumVariance_Cytoplasm_3',1.1814),
('Texture_SumEntropy_Cytoplasm_3',0.9540),
('Texture_Entropy_Cytoplasm_3',1.0677),
('Texture_DifferenceVariance_Cytoplasm_3',0.1314),
('Texture_DifferenceEntropy_Cytoplasm_3',0.4147),
('Texture_InfoMeas1_Cytoplasm_3',-0.4236),
('Texture_InfoMeas2_Cytoplasm_3',0.6609))
for feature_name, value in image_measurements:
py_feature_name = feature_name + "_" + M.H_TO_A[my_angle]
pytm = m.get_current_image_measurement(py_feature_name)
self.assertAlmostEqual(pytm, value,3,
"%s failed. Python=%f, Matlab=%f" %
(feature_name, pytm, value))
Dependencies: Matplotlib, Basemap toolkit, Python Imaging Library
"""
from PIL import Image
from mpl_toolkits.basemap import Basemap
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.image import pil_to_array
plot_name = 'geos_demo.png'
overlay_color = 'black'
# read in jpeg image to rgb array
pilImage = Image.open('200706041200-msg-ch01-SAfrica.jpg')
#data = asarray(pilImage)
data = pil_to_array(pilImage)
data = data[:, :, 0] # get data from first channel in the image
# define data region and projection parameters
ll_lon = 9.74
ll_lat = -35.55
ur_lon = 48.45
ur_lat = 0.2
lon_0 = 0.0
satellite_height = 35785831.0
fig = plt.figure(figsize=(7,7))
ax = fig.add_axes((0.1,0.1,0.8,0.8))
# create Basemap instance for a Geostationary projection.
m = Basemap(projection='geos', lon_0=lon_0, satellite_height=satellite_height,
resolution='l', llcrnrlon=ll_lon, llcrnrlat=ll_lat, urcrnrlon=ur_lon, urcrnrlat=ur_lat)