def process(self, image):
yield 'Start...', image
s_width, s_height = image.size[0] / 2, image.size[1] / 2
simage = image.resize((s_width, s_height))
yield 'posterize...', simage
simage = simage.convert("L")
yield 'posterize 2...', simage
simage = ImageEnhance.Brightness(simage).enhance(2)
simage = ImageOps.equalize(simage)
yield 'posterize 3...', simage
simage = ImageOps.autocontrast(simage, 20)
yield 'posterize 4...', simage
simage = ImageOps.posterize(simage, 1)
colors = []
colors.extend([255] * 128)
colors.extend([0] * 128)
colors.extend([0] * 128)
colors.extend([255] * 128)
colors.extend([0] * 256)
simage = simage.convert("RGB").point(colors)
yield 'Red...', simage
red_img = self._apply_color(simage, 0.82)
yield 'green...', red_img
green_img = self._apply_color(simage, 0.75)
yield 'blue...', green_img
blue_img = self._apply_color(simage, 0)
yield 'yellow...', blue_img
yellow_img = self._apply_color(simage, 0.6)
yield 'Merge...', yellow_img
image = image.copy()
image.paste(red_img, (0, 0, s_width, s_height))
image.paste(green_img, (s_width, 0, s_width * 2, s_height))
image.paste(yellow_img, (0, s_height, s_width, s_height * 2))
image.paste(blue_img, (s_width, s_height, s_width * 2, s_height * 2))
yield 'Done', image
def process(self, image):
yield 'Start...', image
s_width, s_height = image.size[0] / 3, image.size[1] / 3
simage = image.resize((s_width, s_height))
yield 'posterize...', simage
simage = simage.convert("L")
simage = ImageEnhance.Brightness(simage).enhance(2)
simage = ImageOps.equalize(simage)
simage = ImageOps.autocontrast(simage, 20)
simage = ImageOps.posterize(simage, 1)
yield 'colors...', simage
colormap = []
colormap.extend([255] * 128)
colormap.extend([0] * 128)
colormap.extend([0] * 128)
colormap.extend([255] * 128)
colormap.extend([0] * 256)
simage = simage.convert("RGB").point(colormap)
image = image.copy()
for x in xrange(3):
for y in xrange(3):
yield 'Img %d...' % (x * 3 + y + 1), simage
simg = colors.apply_hue_lightness_saturation(simage,
((x * 3 + (2.9 - y)) / 10.), 0, 1, True)
image.paste(simg, (x * s_width, y * s_height,
(x + 1) * s_width, (y + 1) * s_height))
yield 'Done', image
def process(self, image): yield 'Start...', image simage = ImageOps.autocontrast(image, 20) s_width, s_height = image.size[0] / 2, image.size[1] / 2 simage = image.resize((s_width, s_height)) simage = simage.filter(ImageFilter.SMOOTH) yield 'Mask...', simage mask = ImageEnhance.Brightness(simage).enhance(2) mask = ImageOps.posterize(mask, 2) mask = ImageOps.autocontrast(mask, 20) yield 'posterize...', mask simage = ImageEnhance.Brightness(simage).enhance(2) simage = ImageOps.equalize(simage) simage = simage.filter(ImageFilter.SMOOTH_MORE) simage = ImageOps.posterize(simage, 1) yield 'Red...', simage red_img = self._apply_color(simage, 0.25, mask) yield 'green...', red_img green_img = self._apply_color(simage, 0.75, mask) yield 'blue...', green_img blue_img = self._apply_color(simage, 1, mask) yield 'yellow...', blue_img yellow_img = self._apply_color(simage, 0.5, mask) yield 'Merge...', yellow_img image = image.copy() image.paste(red_img, (0, 0, s_width, s_height)) image.paste(green_img, (s_width, 0, s_width * 2, s_height)) image.paste(yellow_img, (0, s_height, s_width, s_height * 2)) image.paste(blue_img, (s_width, s_height, s_width * 2, s_height * 2)) yield 'Done', image
def image_file_features(infile, cl_options):
im = Image.open(infile)
hist_features = {}
if cl_options.histogram:
hist_features = array_to_features(im.histogram(), 'pixhist_')
if cl_options.equalize:
im = ImageOps.grayscale( ImageOps.equalize(im) )
if cl_options.blur:
im_blur = scipy.ndimage.gaussian_filter(im, sigma=cl_options.blur)
im = Image.fromarray(numpy.uint8(im_blur)) # from numpy->Grayscale Image
if cl_options.horiz_ema:
# scale to be independent of image size ?????
im_ema = horiz_exp_mov_avg(im, cl_options.horiz_ema)
im = Image.fromarray(numpy.uint8(im_ema)) # from numpy->Grayscale Image
if cl_options.thumb_size:
thumb_size = (cl_options.thumb_size, cl_options.thumb_size)
im.thumbnail(thumb_size, Image.ANTIALIAS)
if cl_options.thumb_dir:
basename = os.path.basename(infile).split('.')[0]
thumb_file = cl_options.thumb_dir + '/' + basename + '.png'
im.save(thumb_file, "png")
pix_features = {}
if cl_options.pixels:
pix_values = im.getdata() # returns a 1-D/flattened sequence of pixel values
pix_features = array_to_features(pix_values, 'pix_')
return dict(pix_features.items() + hist_features.items())
def equalize(image):
if isinstance(image, pil.Image):
if image.mode in ("F"):
return equalizearray(np.asarray(image))
elif image.mode in ("RBGA"):
image = image.convert("RBG")
return ImageOps.equalize(image)
else:
return equalizearray(image)
def equalize(image):
if isinstance(image, pil.Image):
if image.mode in ("F"):
return equalizearray(np.asarray(image))
elif image.mode in ("RBGA"):
image = image.convert("RBG")
return ImageOps.equalize(image)
else:
return equalizearray(image)
def ConvertToBitifiedImage(file_location,
thumbnail_width=settings.THUMBNAIL_WIDTH,
final_width=settings.FINAL_WIDTH,
bit_depth=settings.BIT_DEPTH):
"""Loads image from filename, generates bitified Image and returns it."""
# Load image from file location
image = Image.open(file_location).convert('RGB')
# If image is JPG/GIF perform extra processing to image
if image.format == 'JPG':
image = ImageOps.autocontrast(image)
elif image.format == 'GIF':
# Need to equalize to change gif to something that can be converted.
image = ImageOps.equalize(image)
# Blur image and reduce number of colors to bit_depth
image = image.convert('P',
palette=Image.ADAPTIVE,
colors=bit_depth).convert('RGB')
processed_image = image.copy()
# Keep only most common color in large area
processed_image = processed_image.filter(ImageFilter.BLUR)
# Low-pass filter
for x in xrange(processed_image.size[0]):
for y in xrange(processed_image.size[1]):
r, g, b = processed_image.getpixel((x, y))
if r < 50 and g < 50 and b < 50:
processed_image.putpixel((x, y), (0, 0, 0))
else:
processed_image.putpixel((x, y), (r, g, b))
# Set final width to at most final_width
width, height = processed_image.size
final_width = min(width, final_width)
# Add border to image before shrinking
processed_image = AddBorderToImage(processed_image)
thumbnail_size = thumbnail_width, thumbnail_width * height/width
final_size = final_width, final_width * height/width
# Shrink image while keeping aspect ratio
processed_image.thumbnail(thumbnail_size, Image.ADAPTIVE)
# Resize back, resulting in pixelized image
processed_image = processed_image.resize(final_size, Image.NEAREST)
# Back to RGB format
processed_image = processed_image.convert('RGB',
palette=Image.ADAPTIVE,
colors=bit_depth)
# Add a random quote to the image
processed_image = AddRandomTextToImage(processed_image)
return processed_image
def ConvertToBitifiedImage(file_location,
thumbnail_width=settings.THUMBNAIL_WIDTH,
final_width=settings.FINAL_WIDTH,
bit_depth=settings.BIT_DEPTH):
"""Loads image from filename, generates bitified Image and returns it."""
# Load image from file location
image = Image.open(file_location).convert('RGB')
# If image is JPG/GIF perform extra processing to image
if image.format == 'JPG':
image = ImageOps.autocontrast(image)
elif image.format == 'GIF':
# Need to equalize to change gif to something that can be converted.
image = ImageOps.equalize(image)
# Blur image and reduce number of colors to bit_depth
image = image.convert('P', palette=Image.ADAPTIVE,
colors=bit_depth).convert('RGB')
processed_image = image.copy()
# Keep only most common color in large area
processed_image = processed_image.filter(ImageFilter.BLUR)
# Low-pass filter
for x in xrange(processed_image.size[0]):
for y in xrange(processed_image.size[1]):
r, g, b = processed_image.getpixel((x, y))
if r < 50 and g < 50 and b < 50:
processed_image.putpixel((x, y), (0, 0, 0))
else:
processed_image.putpixel((x, y), (r, g, b))
# Set final width to at most final_width
width, height = processed_image.size
final_width = min(width, final_width)
# Add border to image before shrinking
processed_image = AddBorderToImage(processed_image)
thumbnail_size = thumbnail_width, thumbnail_width * height / width
final_size = final_width, final_width * height / width
# Shrink image while keeping aspect ratio
processed_image.thumbnail(thumbnail_size, Image.ADAPTIVE)
# Resize back, resulting in pixelized image
processed_image = processed_image.resize(final_size, Image.NEAREST)
# Back to RGB format
processed_image = processed_image.convert('RGB',
palette=Image.ADAPTIVE,
colors=bit_depth)
# Add a random quote to the image
processed_image = AddRandomTextToImage(processed_image)
return processed_image
def equalize(image, amount=100):
image = imtools.convert_safe_mode(image)
if imtools.has_alpha(image):
equalized = imtools.remove_alpha(image)
else:
equalized = image
equalized = ImageOps.equalize(equalized)
if imtools.has_alpha(image):
imtools.put_alpha(equalized, imtools.get_alpha(image))
if amount < 100:
equalized = imtools.blend(image, equalized, amount / 100.0)
return equalized
def equalize(image, amount=100):
image = imtools.convert_safe_mode(image)
if imtools.has_alpha(image):
equalized = imtools.remove_alpha(image)
else:
equalized = image
equalized = ImageOps.equalize(equalized)
if imtools.has_alpha(image):
imtools.put_alpha(equalized, imtools.get_alpha(image))
if amount < 100:
equalized = imtools.blend(image, equalized, amount / 100.0)
return equalized
def PreProcess(im):
im = NMode(im)
im1 = ndimage.grey_erosion(im, size=(10, 10))
scipy.misc.imsave("eroded.jpg", im1)
im1 = Image.open("eroded.jpg")
im = ImageOps.equalize(im, 0)
im = ImageChops.difference(im1, im)
#print ("image height %d and width %d\n"%(imh,imw))
im = GBinarization(im) #binarize the image
return im
def PreProcess(im):
im=NMode(im)
im1 = ndimage.grey_erosion(im, size=(10,10))
scipy.misc.imsave("eroded.jpg",im1)
im1= Image.open("eroded.jpg")
im=ImageOps.equalize(im,0)
im=ImageChops.difference(im1, im)
#print ("image height %d and width %d\n"%(imh,imw))
im=GBinarization(im)#binarize the image
return im
def GenEqual(file, label):
'''
Generate Equalization feature
There is a bug: when the total pixel count less than 255,ImageOps.equalize take no effect
'''
noext, ext = os.path.splitext(file)
if(ext == '.png'):
im = Image.open(file)
im = im.convert('L')
im = ImageOps.equalize(im)
data = im.getdata()
for pixel in data:
outputFile.write(str(pixel)+' ')
outputFile.write(str(label)+ '\n')
def GenEqual(file, label):
'''
Generate Equalization feature
There is a bug: when the total pixel count less than 255,ImageOps.equalize take no effect
'''
noext, ext = os.path.splitext(file)
if (ext == '.png'):
im = Image.open(file)
im = im.convert('L')
im = ImageOps.equalize(im)
data = im.getdata()
for pixel in data:
outputFile.write(str(pixel) + ' ')
outputFile.write(str(label) + '\n')
def dem2img(self, path, file):
# Source LIDAR DEM file
source = path + file
# Load the ASCII DEM into a numpy array
arr = np.loadtxt(source, skiprows=6)
# Convert the numpy array to a PIL image
im = Image.fromarray(arr).convert('L')
# Enhance the image
im = ImageOps.equalize(im)
im = ImageOps.autocontrast(im)
# Begin building our color ramp
palette = []
# Hue, Saturaction, Value
# color space
h = .67
s = 1
v = 1
step = h / 256.0
for i in range(256):
rp, gp, bp = colorsys.hsv_to_rgb(h, s, v)
r = int(rp * 255)
g = int(gp * 255)
b = int(bp * 255)
palette.extend([r, g, b])
h -= step
# Apply the palette to the image
im.putpalette(palette)
im = im.convert('RGB')
# Output image file
img_path = os.path.dirname(os.getcwd(
)) + '/webGIS/static/showLidar/images/' + file.split('.')[0] + '.jpg'
# Save the image
im.save(img_path)
image = '/static/showLidar/images/' + file.split('.')[0] + '.jpg'
return image
def showResult(n,result, dims, k=2):
i = Image.new("RGB", dims)
i.putdata(result, int(255 / (k - 1)))
i = ImageOps.equalize(i)
new = pil_to_pygame_img(i)
screen = pygame.display.get_surface()
if not screen.get_width == dims[1]:
x=dims[0]
y=dims[1]
initScreen(x,y)
screen.blit(new,(0,0))
if pygame.font:
font = pygame.font.Font(None, 36)
ts = 'Rule ' + str(n)
text = font.render(ts, 1, (255, 255, 10))
textpos = text.get_rect()
textpos.centerx = screen.get_rect().centery
screen.blit(text, textpos)
pygame.display.flip()
def normalize(image):
image = image.filter(ImageFilter.BLUR)
picture = ImageChops.blend(ImageOps.equalize(image), image, .5)
return ImageChops.multiply(picture, picture)
import colorsys
# Source LIDAR DEM file
source = "lidar.asc"
# Output image file
target = "lidar.bmp"
# Load the ASCII DEM into a numpy array
arr = np.loadtxt(source, skiprows=6)
# Convert the numpy array to a PIL image
im = Image.fromarray(arr).convert('L')
# Enhance the image
im = ImageOps.equalize(im)
im = ImageOps.autocontrast(im)
# Begin building our color ramp
palette = []
# Hue, Saturaction, Value
# color space
h = .67
s = 1
v = 1
# We'll step through colors from:
# blue-green-yellow-orange-red.
# Blue=low elevation, Red=high-elevation
step = h / 256.0
def augment(self, image):
return ImageOps.equalize(image)
def process(self, _edObject=None):
"""
This is the main method of the plugin, it does the job, not the EDNA name conversions which were done in the preprocess.
"""
EDPluginExec.process(self)
EDVerbose.DEBUG("EDPluginExecThumbnailv10.process")
# try:
# except Exception:
# edfImage = EDF(self.inputFilename)
# self.npaImage = edfImage.GetData(0)
# Read the image using FABIO
isRGB = False
pilOutputImage = None
if self.inputFilename is not None:
try:
fabioImage = openimage(self.inputFilename)
self.npaImage = fabioImage.data
except Exception:
pilInputImage = Image.open(self.inputFilename)
x, y = pilInputImage.size
ImageFile.MAXBLOCK = x * y
if pilInputImage.mode == "1":
self.npaImage = numpy.asarray(pilInputImage).astype("uint8")
isRGB = False
elif pilInputImage.mode == "F":
self.npaImage = numpy.asarray(pilInputImage)
isRGB = False
elif pilInputImage.mode == "L":
self.npaImage = numpy.asarray(pilInputImage)
isRGB = False
elif pilInputImage.mode == "P":
self.npaImage = numpy.asarray(pilInputImage.convert("RGB"))
isRGB = True
elif pilInputImage.mode == "RGB":
self.npaImage = numpy.asarray(pilInputImage)
isRGB = True
elif pilInputImage.mode == "CMJK":
self.npaImage = numpy.asarray(pilInputImage.convert("RGB"))
isRGB = True
dtype = self.npaImage.dtype
NPAImageFloat = None
# crop border
if len(self.cropBorders) > 0:
if len(self.cropBorders) == 1:
crop0 = self.cropBorders[0]
crop1 = self.cropBorders[0]
else:
crop0 = self.cropBorders[0]
crop1 = self.cropBorders[1]
if isRGB:
self.npaImage = self.npaImage[crop0:-crop0, crop1:crop1, :]
else:
self.npaImage = self.npaImage[crop0:-crop0, crop1:crop1]
# Set maxima and minima
if (self.minLevelUnit is not None) or (self.maxLevelUnit is not None):
sortedArray = self.npaImage.flatten()
sortedArray.sort()
if self.minLevel is not None:
self.normalize = True
if isRGB:
EDVerbose.warning("It is not allowed to set Min with RGB data")
else:
if self.minLevelUnit in ["%", "percent"]:
self.minLevel = sortedArray[int(round(float(self.minLevel) * sortedArray.size / 100.0))]
if isinstance(self.npaImage[0, 0], int):
self.npaImage = numpy.maximum(self.npaImage, int(self.minLevel) * numpy.ones_like(self.npaImage))
else:
self.npaImage = numpy.maximum(self.npaImage, self.minLevel * numpy.ones_like(self.npaImage))
if self.maxLevel is not None:
self.normalize = True
if isRGB:
EDVerbose.warning("It is not allowed to set Max with RGB data")
else:
if self.maxLevelUnit in ["%", "percent"]:
self.maxLevel = sortedArray[int(round(float(self.maxLevel) * sortedArray.size / 100.0))]
if isinstance(self.npaImage[0, 0], int):
self.npaImage = numpy.minimum(self.npaImage, int(self.maxLevel) * numpy.ones_like(self.npaImage))
else:
self.npaImage = numpy.minimum(self.npaImage, self.maxLevel * numpy.ones_like(self.npaImage))
# Scipy filters come here:
if len(self.gaussianBlur) > 0:
if len(self.gaussianBlur) == 1 :
kernel = (self.gaussianBlur[0], self.gaussianBlur[0])
else:
kernel = (self.gaussianBlur[0], self.gaussianBlur[1])
if isRGB:
kernel = (kernel[0], kernel[1], 0)
self.npaImage = scipy.ndimage.gaussian_filter(self.npaImage, kernel)
if len(self.dilatation) > 0:
if len(self.dilatation) == 1:
kernel = (self.dilatation[0], self.dilatation[0])
else:
kernel = (self.dilatation[0], self.dilatation[1])
if isRGB:
kernel = (kernel[0], kernel[1], 0)
self.npaImage = scipy.ndimage.morphology.grey_dilation(self.npaImage, kernel)
#Normalization ; equalization
if (self.normalize is True) or (self.equalize is True):
if isRGB is True:
self.npaImage = numpy.asarray(ImageOps.equalize(Image.fromarray(self.npaImage)))
else:
EDVerbose.DEBUG("EDPluginExecThumbnailv10: Normalization")
vmin = self.npaImage.min()
vmax = self.npaImage.max()
NPAImageFloat = (self.npaImage.astype(numpy.float32) - float(vmin)) / (float(vmax) - float(vmin))
if (self.equalize == True):
nbr_bins = 64
NPAImageFloatFlat = NPAImageFloat.flatten()
imhist, bins = numpy.histogram(NPAImageFloatFlat, nbr_bins, normed=True) #get image histogram
cdf = imhist.cumsum() #cumulative distribution function
ncdf = cdf / cdf[-1] #normalized cumulative distribution function
# print ncdf
NPAImageFloat2Flat = numpy.interp(NPAImageFloatFlat, bins, [0] + ncdf.tolist())
NPAImageFloat = NPAImageFloat2Flat.reshape(NPAImageFloat.shape)
EDVerbose.DEBUG("Equalize: min= %f, max= %f" % (NPAImageFloat.min(), NPAImageFloat.max()))
#Gamma and logarithm scale
if ((self.log is True) or (self.gamma != 1)) and (NPAImageFloat is None): # then we need the array in float
if dtype == numpy.uint8:
NPAImageFloat = self.npaImage.astype(numpy.float32) / 255.0
elif dtype == numpy.uint16:
NPAImageFloat = self.npaImage.astype(numpy.float32) / 65535.0
else:
NPAImageFloat = self.npaImage.astype(numpy.float32)
if self.log is True:
NPAImageFloat = numpy.log(1 - NPAImageFloat.min() + NPAImageFloat)
vmin = NPAImageFloat.min()
vmax = NPAImageFloat.max()
NPAImageFloat = (NPAImageFloat - vmin) / (vmax - vmin)
if self.gamma != 1:
if dtype not in [numpy.uint8, numpy.uint16]:
vmin = NPAImageFloat.min()
vmax = NPAImageFloat.max()
NPAImageFloat = (NPAImageFloat - vmin) / (vmax - vmin)
NPAImageInt = (255.0 * (NPAImageFloat ** self.gamma)).astype("uint8")
else: #if (self.gamma == 1):
if NPAImageFloat is None:
if dtype == numpy.uint8:
NPAImageInt = self.npaImage
elif dtype == numpy.uint16:
NPAImageInt = (self.npaImage / 256).astype(numpy.uint8)
else: #for float or a signed integer
vmin = self.npaImage.min()
vmax = self.npaImage.max()
NPAImageInt = ((self.npaImage.astype(numpy.float32) - vmin) / (vmax - vmin) * 255.0).astype(numpy.uint8)
else:
vmin = NPAImageFloat.min()
vmax = NPAImageFloat.max()
EDVerbose.DEBUG("EDPluginExecThumbnailv10: NPAImageFloat => NPAImageInt min=%s max =%s" % (vmin, vmax))
NPAImageInt = ((NPAImageFloat - vmin) * 255.0 / (vmax - vmin)).astype(numpy.uint8)
#COnversion back to PIL mode
if isRGB is True:
pilOutputImage = Image.fromarray(NPAImageInt, 'RGB')
else:
pilOutputImage = Image.fromarray(NPAImageInt, 'L')
if (self.autocontrast is not None):
pilOutputImage = ImageOps.autocontrast(pilOutputImage, self.autocontrast)
if (self.width is not None) or (self.height is not None):
if (self.width > 0) and (self.height > 0):
if self.keepRatio is True:
# PIL takes care of the ratio
pilOutputImage.thumbnail((self.width, self.height), Image.ANTIALIAS)
else:
pilOutputImage = pilOutputImage.resize((self.width, self.height), Image.ANTIALIAS)
else:
if self.width is None:
pilOutputImage.thumbnail((self.height, self.height), Image.ANTIALIAS)
elif self.height is None:
pilOutputImage.thumbnail((self.width, self.width), Image.ANTIALIAS)
if self.invert == True:
pilOutputImage = ImageOps.invert(pilOutputImage)
if self.colorize == True:
pilOutputImage.putpalette(EDPluginExecThumbnailv10.getPalette())
pilOutputImage = pilOutputImage.convert("RGB")
self.synchronizeOn()
if self.format == "jpg":
self.width, self.height = pilOutputImage.size
if self.width * self.height > ImageFile.MAXBLOCK:
ImageFile.MAXBLOCK = self.width * self.height
try:
pilOutputImage.save(self.output, "JPEG", quality=85, optimize=True)
except TypeError:
pilOutputImage.save(self.output)
else:
pilOutputImage.save(self.output)
self.synchronizeOff()
def process(self, _edObject=None):
"""
This is the main method of the plugin, it does the job, not the EDNA name conversions which were done in the preprocess.
"""
EDPluginExec.process(self)
EDVerbose.DEBUG("EDPluginExecThumbnailv10.process")
# try:
# except Exception:
# edfImage = EDF(self.inputFilename)
# self.npaImage = edfImage.GetData(0)
# Read the image using FABIO
isRGB = False
pilOutputImage = None
if self.inputFilename is not None:
try:
fabioImage = openimage(self.inputFilename)
self.npaImage = fabioImage.data
except Exception:
pilInputImage = Image.open(self.inputFilename)
x, y = pilInputImage.size
ImageFile.MAXBLOCK = x * y
if pilInputImage.mode == "1":
self.npaImage = numpy.asarray(pilInputImage).astype(
"uint8")
isRGB = False
elif pilInputImage.mode == "F":
self.npaImage = numpy.asarray(pilInputImage)
isRGB = False
elif pilInputImage.mode == "L":
self.npaImage = numpy.asarray(pilInputImage)
isRGB = False
elif pilInputImage.mode == "P":
self.npaImage = numpy.asarray(pilInputImage.convert("RGB"))
isRGB = True
elif pilInputImage.mode == "RGB":
self.npaImage = numpy.asarray(pilInputImage)
isRGB = True
elif pilInputImage.mode == "CMJK":
self.npaImage = numpy.asarray(pilInputImage.convert("RGB"))
isRGB = True
dtype = self.npaImage.dtype
NPAImageFloat = None
# crop border
if len(self.cropBorders) > 0:
if len(self.cropBorders) == 1:
crop0 = self.cropBorders[0]
crop1 = self.cropBorders[0]
else:
crop0 = self.cropBorders[0]
crop1 = self.cropBorders[1]
if isRGB:
self.npaImage = self.npaImage[crop0:-crop0, crop1:crop1, :]
else:
self.npaImage = self.npaImage[crop0:-crop0, crop1:crop1]
# Set maxima and minima
if (self.minLevelUnit is not None) or (self.maxLevelUnit is not None):
sortedArray = self.npaImage.flatten()
sortedArray.sort()
if self.minLevel is not None:
self.normalize = True
if isRGB:
EDVerbose.warning(
"It is not allowed to set Min with RGB data")
else:
if self.minLevelUnit in ["%", "percent"]:
self.minLevel = sortedArray[int(
round(float(self.minLevel) * sortedArray.size /
100.0))]
if isinstance(self.npaImage[0, 0], int):
self.npaImage = numpy.maximum(
self.npaImage,
int(self.minLevel) * numpy.ones_like(self.npaImage))
else:
self.npaImage = numpy.maximum(
self.npaImage,
self.minLevel * numpy.ones_like(self.npaImage))
if self.maxLevel is not None:
self.normalize = True
if isRGB:
EDVerbose.warning(
"It is not allowed to set Max with RGB data")
else:
if self.maxLevelUnit in ["%", "percent"]:
self.maxLevel = sortedArray[int(
round(float(self.maxLevel) * sortedArray.size /
100.0))]
if isinstance(self.npaImage[0, 0], int):
self.npaImage = numpy.minimum(
self.npaImage,
int(self.maxLevel) * numpy.ones_like(self.npaImage))
else:
self.npaImage = numpy.minimum(
self.npaImage,
self.maxLevel * numpy.ones_like(self.npaImage))
# Scipy filters come here:
if len(self.gaussianBlur) > 0:
if len(self.gaussianBlur) == 1:
kernel = (self.gaussianBlur[0], self.gaussianBlur[0])
else:
kernel = (self.gaussianBlur[0], self.gaussianBlur[1])
if isRGB:
kernel = (kernel[0], kernel[1], 0)
self.npaImage = scipy.ndimage.gaussian_filter(
self.npaImage, kernel)
if len(self.dilatation) > 0:
if len(self.dilatation) == 1:
kernel = (self.dilatation[0], self.dilatation[0])
else:
kernel = (self.dilatation[0], self.dilatation[1])
if isRGB:
kernel = (kernel[0], kernel[1], 0)
self.npaImage = scipy.ndimage.morphology.grey_dilation(
self.npaImage, kernel)
#Normalization ; equalization
if (self.normalize is True) or (self.equalize is True):
if isRGB is True:
self.npaImage = numpy.asarray(
ImageOps.equalize(Image.fromarray(self.npaImage)))
else:
EDVerbose.DEBUG("EDPluginExecThumbnailv10: Normalization")
vmin = self.npaImage.min()
vmax = self.npaImage.max()
NPAImageFloat = (self.npaImage.astype(numpy.float32) -
float(vmin)) / (float(vmax) - float(vmin))
if (self.equalize == True):
nbr_bins = 64
NPAImageFloatFlat = NPAImageFloat.flatten()
imhist, bins = numpy.histogram(
NPAImageFloatFlat, nbr_bins,
normed=True) #get image histogram
cdf = imhist.cumsum() #cumulative distribution function
ncdf = cdf / cdf[
-1] #normalized cumulative distribution function
# print ncdf
NPAImageFloat2Flat = numpy.interp(NPAImageFloatFlat, bins,
[0] + ncdf.tolist())
NPAImageFloat = NPAImageFloat2Flat.reshape(
NPAImageFloat.shape)
EDVerbose.DEBUG("Equalize: min= %f, max= %f" %
(NPAImageFloat.min(), NPAImageFloat.max()))
#Gamma and logarithm scale
if ((self.log is True) or (self.gamma != 1)) and (
NPAImageFloat is None): # then we need the array in float
if dtype == numpy.uint8:
NPAImageFloat = self.npaImage.astype(numpy.float32) / 255.0
elif dtype == numpy.uint16:
NPAImageFloat = self.npaImage.astype(numpy.float32) / 65535.0
else:
NPAImageFloat = self.npaImage.astype(numpy.float32)
if self.log is True:
NPAImageFloat = numpy.log(1 - NPAImageFloat.min() + NPAImageFloat)
vmin = NPAImageFloat.min()
vmax = NPAImageFloat.max()
NPAImageFloat = (NPAImageFloat - vmin) / (vmax - vmin)
if self.gamma != 1:
if dtype not in [numpy.uint8, numpy.uint16]:
vmin = NPAImageFloat.min()
vmax = NPAImageFloat.max()
NPAImageFloat = (NPAImageFloat - vmin) / (vmax - vmin)
NPAImageInt = (255.0 * (NPAImageFloat**self.gamma)).astype("uint8")
else: #if (self.gamma == 1):
if NPAImageFloat is None:
if dtype == numpy.uint8:
NPAImageInt = self.npaImage
elif dtype == numpy.uint16:
NPAImageInt = (self.npaImage / 256).astype(numpy.uint8)
else: #for float or a signed integer
vmin = self.npaImage.min()
vmax = self.npaImage.max()
NPAImageInt = (
(self.npaImage.astype(numpy.float32) - vmin) /
(vmax - vmin) * 255.0).astype(numpy.uint8)
else:
vmin = NPAImageFloat.min()
vmax = NPAImageFloat.max()
EDVerbose.DEBUG(
"EDPluginExecThumbnailv10: NPAImageFloat => NPAImageInt min=%s max =%s"
% (vmin, vmax))
NPAImageInt = ((NPAImageFloat - vmin) * 255.0 /
(vmax - vmin)).astype(numpy.uint8)
#COnversion back to PIL mode
if isRGB is True:
pilOutputImage = Image.fromarray(NPAImageInt, 'RGB')
else:
pilOutputImage = Image.fromarray(NPAImageInt, 'L')
if (self.autocontrast is not None):
pilOutputImage = ImageOps.autocontrast(pilOutputImage,
self.autocontrast)
if (self.width is not None) or (self.height is not None):
if (self.width > 0) and (self.height > 0):
if self.keepRatio is True:
# PIL takes care of the ratio
pilOutputImage.thumbnail((self.width, self.height),
Image.ANTIALIAS)
else:
pilOutputImage = pilOutputImage.resize(
(self.width, self.height), Image.ANTIALIAS)
else:
if self.width is None:
pilOutputImage.thumbnail((self.height, self.height),
Image.ANTIALIAS)
elif self.height is None:
pilOutputImage.thumbnail((self.width, self.width),
Image.ANTIALIAS)
if self.invert == True:
pilOutputImage = ImageOps.invert(pilOutputImage)
if self.colorize == True:
pilOutputImage.putpalette(EDPluginExecThumbnailv10.getPalette())
pilOutputImage = pilOutputImage.convert("RGB")
self.synchronizeOn()
if self.format == "jpg":
self.width, self.height = pilOutputImage.size
if self.width * self.height > ImageFile.MAXBLOCK:
ImageFile.MAXBLOCK = self.width * self.height
try:
pilOutputImage.save(self.output,
"JPEG",
quality=85,
optimize=True)
except TypeError:
pilOutputImage.save(self.output)
else:
pilOutputImage.save(self.output)
self.synchronizeOff()
def normalize(image):
image = image.filter(ImageFilter.BLUR)
picture = ImageChops.blend(ImageOps.equalize(image), image, .5)
return ImageChops.multiply(picture, picture)
def augment(self, image):
return ImageOps.equalize(image)
def raster(result_img, filt, bbox = (-999,-999,999,9999), srs="EPSG:3857"):
"""
Applies various filters to image.
"""
for ff in filt:
if ff.split(":") == [ff,]:
if ff == "bw":
result_img = result_img.convert("L")
result_img = result_img.convert("RGBA")
if ff == "contour":
result_img = result_img.filter(ImageFilter.CONTOUR)
if ff == "median":
result_img = result_img.filter(ImageFilter.MedianFilter(5))
if ff == "blur":
result_img = result_img.filter(ImageFilter.BLUR)
if ff == "edge":
result_img = result_img.filter(ImageFilter.EDGE_ENHANCE)
if ff == "equalize":
result_img = result_img.convert("RGB")
result_img = ImageOps.equalize(result_img)
result_img = result_img.convert("RGBA")
if ff == "autocontrast":
result_img = result_img.convert("RGB")
result_img = ImageOps.autocontrast(result_img)
result_img = result_img.convert("RGBA")
if ff == "swaprb":
ii = result_img.split()
result_img = Image.merge("RGBA", (ii[2],ii[1],ii[0],ii[3]))
else:
ff, tts = ff.split(":")
try:
tt = float(tts)
except:
tt = 1
pass
if ff == "brightness":
enhancer = ImageEnhance.Brightness(result_img)
result_img = enhancer.enhance(tt)
if ff == "contrast":
enhancer = ImageEnhance.Contrast(result_img)
result_img = enhancer.enhance(tt)
if ff == "sharpness":
enhancer = ImageEnhance.Sharpness(result_img)
result_img = enhancer.enhance(tt)
if ff == "autocontrast":
result_img = result_img.convert("RGB")
result_img = ImageOps.autocontrast(result_img, tt)
result_img = result_img.convert("RGBA")
if ff == "fusion" and NUMPY_AVAILABLE:
pix = numpy.array(result_img, dtype=int)
a,b = result_img.size
pan_img = getimg (bbox, srs, [b, a], config.layers[tts], datetime.datetime.now(), [])
pan_img = pan_img.convert("L")
print pix.dtype
print pix[...,1].shape
pan = numpy.array(pan_img)
pix[...,0] = pix[...,0]*pan/(pix[...,0] + pix[...,1] + pix[...,2])
pix[...,1] = pix[...,1]*pan/(pix[...,0] + pix[...,1] + pix[...,2])
pix[...,2] = pix[...,2]*pan/(pix[...,0] + pix[...,1] + pix[...,2])
print pix.shape
result_img = Image.fromarray(numpy.uint8(pix))
result_img = result_img.convert("RGBA")
return result_img
#im8 = im.filter(ImageFilter.MedianFilter(size=3)) #im8.show() #mode filter #im8 = im.filter(ImageFilter.ModeFilter(size=3)) #im8.show() #EDGE_ENHANCEfilter #im8 = im.filter(ImageFilter.EDGE_ENHANCE) #im8.show() #greyscale the image im8 = ImageOps.grayscale(im) im8.show() #equalize the image im8 = ImageOps.equalize(im6) im8.show() #get image stats im_stat = ImageStat.Stat(im) print "stat.extrema = ", im_stat.extrema print "stat.count = ", im_stat.count print "stat.sum = ", im_stat.sum print "stat.sum2 = ", im_stat.sum2 print "stat.mean = ", im_stat.mean print "stat.median = ", im_stat.median print "stat.rms = ", im_stat.rms print "stat.var = ", im_stat.var print "stat.stddev = ", im_stat.stddev