def f2():
import pgm
#
# imgs is the image to be modified
# imgt is the reference image
#
imgarray,w,h = pgm.pgmread('histo_inp_image.pgm')
#imgarray,w,h = pgm.pgmread('histo_ref_image.pgm')
#imgarray,w,h = pgm.pgmread('histo_inp_image2.pgm')
#imgarray,w,h = pgm.pgmread('histo_ref_image2.pgm')
#imgarray,w,h = pgm.pgmread('lena.pgm')
#imgarray,w,h = pgm.pgmread('fourier_transform.pgm')
#imgarray,w,h = pgm.pgmread('Fourier.pgm')
imgs = Img.Img(imgarray)
#
#imgarray,w,h = pgm.pgmread('histo_inp_image.pgm')
imgarray,w,h = pgm.pgmread('histo_ref_image.pgm')
#imgarray,w,h = pgm.pgmread('histo_inp_image2.pgm')
#imgarray,w,h = pgm.pgmread('histo_ref_image2.pgm')
#imgarray,w,h = pgm.pgmread('lena.pgm')
#imgarray,w,h = pgm.pgmread('fourier_transform.pgm')
#imgarray,w,h = pgm.pgmread('Fourier.pgm')
imgt = Img.Img(imgarray)
#
imgt1 = imgs.modify_hist(imgt)
plot_fig(imgs.pix, 'Original image')
plot_fig(imgt.pix, 'Reference image')
plot_fig(imgt1.pix, 'Modified original image')
stem_hist(imgs, 'Original image')
stem_hist(imgt, 'Reference image')
stem_hist(imgt1, 'Modified original image')
print np.sum((imgs.pix - imgt1.pix)**2)
def f1():
""" Convolves image and kernel """
(imgarr, w, h) = pgm.pgmread('cameraman.pgm')
imgs = Img.Img(imgarr)
k1 = Img.create_gauss_kernel(Img.find_gauss_support(0.8))
k2 = Img.create_gauss_kernel(Img.find_gauss_support(1.2))
k3 = Img.create_gauss_kernel(Img.find_gauss_support(1.6))
imgt1 = imgs.conv(k1)
imgt2 = imgs.conv(k2)
imgt3 = imgs.conv(k3)
# print imgt1.col, imgt1.row
# print imgt2.col, imgt2.row
# print imgt3.col, imgt3.row
plt.figure(1)
plt.imshow(imgs.pix, cmap=cm.gray)
plt.title('Original image')
plt.figure(2)
plt.imshow(imgt1.pix, cmap=cm.gray)
plt.title('Img convolved with Gaussian kernel' + str(shape(k1)))
plt.figure(3)
plt.imshow(imgt2.pix, cmap=cm.gray)
plt.title('Img convolved with Gaussian kernel' + str(shape(k2)))
plt.figure(4)
plt.imshow(imgt3.pix, cmap=cm.gray)
plt.title('Img convolved with Gaussian kernel' + str(shape(k3)))
plt.show()
def f1():
import pgm
#imgarray,w,h = pgm.pgmread('histo_inp_image.pgm')
#imgarray,w,h = pgm.pgmread('histo_ref_image.pgm')
#imgarray,w,h = pgm.pgmread('histo_inp_image2.pgm')
#imgarray,w,h = pgm.pgmread('histo_ref_image2.pgm')
imgarray,w,h = pgm.pgmread('lena.pgm')
#imgarray,w,h = pgm.pgmread('fourier_transform.pgm')
imgs = Img.Img(imgarray)
imgt = imgs.add_salt_pepper(0.65)
imgt1 = imgt.median_filter(np.ones((3,3)))
imgt1 = imgt1.median_filter(np.ones((5,5)))
#imgt1 = imgt.median_filter(np.ones((7,7)))
#imgt1 = imgt.median_filter(np.array([[2,2,2],[2,1,2],[2,2,2]]))
k = Img.create_gauss_kernel(Img.find_gauss_support(0.75), 0.75)
imgt2 = imgt.conv(k)
plot_fig(imgs.pix, 'Original image')
plot_fig(imgt.pix, 'Added salt and pepper noise')
plot_fig(imgt1.pix, 'Median filtered image')
plot_fig(imgt2.pix, 'Gaussian LP filtered image')
#plot_hist(imgs, 'Original image')
#plot_hist(imgt, 'Added salt and pepper noise')
#plot_hist(imgt1, 'Median filtered image')
#plot_hist(imgt2, 'Gaussian LP filtered image')
print np.sum((imgs.pix - imgt1.pix)**2)
def calc_sml(file1, file2, q):
""" Calculates SML for file1 frames with q(defined as in class),
and store as mat file file2
"""
stack = get_stack(file1)
stackOut = dict() # Define output stack as dict
for key in sorted(stack.keys()):
print key
imgOut = Img.Img(stack[key]).sum_mod_lap(q) # create a new Img class
stackOut[key] = imgOut.pix # Add array to output stack
io.savemat(file2, stackOut)
def Preprocessing(img, option='edge'):
imdata = Img(img)
if (option == 'raw' or option == 'RAW' or option == 'Raw'):
imdata.setGrayscaleBuffer()
imdata.znorm()
imdata.normalize()
elif (option == 'edge' or option == 'EDGE' or option == 'Edge'):
imdata.setGrayscaleBuffer()
imdata.znorm()
imdata = ExtractEdge(imdata)
#imdata.display()
return imdata
def f3():
import pgm
imgarray,w,h = pgm.pgmread('histo_inp_image.pgm')
#imgarray,w,h = pgm.pgmread('histo_ref_image.pgm')
#imgarray,w,h = pgm.pgmread('histo_inp_image2.pgm')
#imgarray,w,h = pgm.pgmread('histo_ref_image2.pgm')
#imgarray,w,h = pgm.pgmread('lena.pgm')
#imgarray,w,h = pgm.pgmread('fourier_transform.pgm')
#imgarray,w,h = pgm.pgmread('Fourier.pgm')
#imgarray,w,h = pgm.pgmread('academy.pgm')
imgs = Img.Img(imgarray)
imgt = imgs.hist_eq()
plot_fig(imgs.pix, 'Original image')
plot_fig(imgt.pix, 'Histogram equalized image')
stem_hist(imgs, 'Original image')
stem_hist(imgt, 'Histogram equalized image')
print np.sum((imgs.pix - imgt.pix)**2)
def f2():
""" Scale image """
factor = (0.8, 0.8)
(imgarr, w, h) = pgm.pgmread('cameraman.pgm')
imgs = Img.Img(imgarr)
imgt1 = imgs.scale(factor)
imgt2 = imgs.scale(factor, type='bilinear')
plt.figure(1)
plt.imshow(imgs.pix, cmap=cm.gray)
plt.title('Original image')
plt.figure(2)
plt.imshow(imgt1.pix, cmap=cm.gray)
plt.title('Scale factor = ' + str(factor))
plt.figure(3)
plt.imshow(imgt2.pix, cmap=cm.gray)
plt.title('Scale factor = ' + str(factor) +
'\nwith bilinear interpolation')
plt.show()
def __init__(self, bs, n=None):
print('Building computational graph...')
self.bs = bs
self.path = filepath
self.I = Img()
self.n = n
self.I.trainlist = self.I.trainlist[0:n]
x = T.tensor4('inputs')
y = T.tensor4('targets')
loss = self.build(x, y)
updat = Tool.adam(loss, self.params)
self.train = theano.function([x, y], loss, updates=updat)
self.predict = theano.function([x], self.Y)
print('Computational graph built.')
def GetFaceChip(im, src_leye, src_reye, dst_leye, dst_reye, dst_size, channel='gray'):
imdata = Img(im)
if (channel=='gray' or channel=='Gray' or channel=='GRAY'):
imdata.setGrayscaleBuffer()
elif (channel=='RGB'):
imdata.setRGBBuffer()
elif (channel=='red' or channel=='Red' or channel=='RED'):
imdata.setRedBuffer()
elif (channel=='I'):
imdata.setChrominanceBuffer()
else:
raise TypeError("Unsuppoted Color Channel: %s ..."%channel)
## smooth the image before applying affine transformation
imdata = FaceSmoothing(imdata, src_leye, src_reye)
imdata = AffineFromPoints(imdata, src_leye, src_reye, dst_leye, dst_reye, dst_size)
return imdata
def sum_mod_lap_test():
imgs = Img.Img(np.array([[1, 2, 3], [0, 1, 5], [6, 3, 9]]))
imgt = imgs.sum_mod_lap(1)
print imgs.pix
print imgt.pix
def g1():
import pgm
imgarray,w,h = pgm.pgmread('test.pgm')
imgs = Img.Img(imgarray)
imgt1 = imgs.median_filter(np.ones((3,3)))
pgm.pgmwrite(imgt1.pix, 'test2.pgm')
imgarray = img.calc_hist()
plt.figure()
plt.plot(imgarray)
plt.title(plotTitle)
def plot_fig(imgarr, plotTitle):
plt.figure()
plt.imshow(imgarr, cmap=cm.gray)
plt.title(plotTitle)
if __name__ == '__main__':
import pgm
(imgarray, w, h) = pgm.pgmread('academy.pgm')
img = Img.Img(imgarray)
# plot_fig(np.int32(img.pix), 'Original mage')
# plot_hist(img, 'Original image')
imgarray = add_noise(imgarray)
imgs = Img.Img(imgarray)
plot_fig(imgs.pix, 'Image with additive Gaussian with sigma=5')
pgm.pgmwrite(np.int32(imgs.pix), 'academy_with_noise.pgm')
# plot_hist(img, 'Noise image')
##
# f1(imgs, 1, 1)
# f1(imgs, 1, 10)
# f1(imgs, 1, 100)
# f1(imgs, 1, 300)
##
# f1(imgs, 3, 1)
# f1(imgs, 3, 10)
image_no = 6
#Insert the 5 epoc numbers you'd like to compare.
model_trained = [5, 10, 15, 20, 25]
for imgno in range(9):
image_no = imgno
reconstructions = []
for i in range(5):
M = Model(batch_size)
M.__load__(model_trained[i])
M.Generate(batch_type, batch_no)
reconstructions.append(M.valid_recon[image_no])
I = Img()
curbatch, curcrop = I.load_batch(batch_size, batch_no, batch_type)
curcropped = np.copy(curbatch)
curbatch[:, :, 16:48, 16:48] = curcrop
#I.plotandcompare(M.valid_recon[image_no], curbatch[image_no])
for i in range(len(reconstructions)):
reconstructions[i] = reconstructions[i].transpose(1, 2, 0)
original = curbatch[image_no].transpose(1, 2, 0)
challenge = curcropped[image_no].transpose(1, 2, 0)
fig = plt.figure(figsize=(17, 3))
ax1 = fig.add_subplot(171)
ax1.imshow(challenge)
def DisplayImg(img, m, n):
pic = Img(img, m, n)
pic.display()
def __init__(self,root):
filename = filedialog.askopenfilename(initialdir="/", title="Selectionnez une image",
filetypes=(("jpeg files", "*.jpg"), ("all files", "*.*")))
Im = Img(filename)
trait = tk.Toplevel(root)
trait.title("Traitements d'images")
trait.geometry("%dx%d" % (trait.winfo_screenwidth(), trait.winfo_screenheight()))
## Déclaration des éléments graphique de type menu
menubar = Menu(trait)
trait['menu'] = menubar
Fichier = Menu(menubar)
Traitement = Menu(menubar)
Couleur = Menu(menubar)
Texture = Menu(menubar)
Contour = Menu(menubar)
Aide = Menu(menubar)
Change_palette = Menu(Couleur)
Change_color = Menu(Couleur)
can2 = Canvas(trait, bg="white")
can2.pack(side=tk.TOP, fill=BOTH, expand=True)
can2.image = Im.getImagetk()
can2.create_image(trait.winfo_screenwidth()/4, 0, anchor=NW, image=Im.getImagetk())
## Ajout d'un onglet Fichier
menubar.add_cascade(label="Fichier", menu=Fichier)
Fichier.add_command(label="Enregistrer_sous", command=exit)
Fichier.add_command(label="Quitter", command=exit)
## Ajout d'un onglet Traitement
menubar.add_cascade(label="Traitement", menu=Traitement)
Traitement.add_command(label="Egalisation d'histogramme", command=Im.egalisation)
Traitement.add_command(label="Reconaissance de lettres", command=exit)
Traitement.add_command(label="Segmentation", command=Im.segmentation)
## Ajout d'un onglet Couleur
menubar.add_cascade(label="Couleur", menu=Couleur)
Couleur.add_command(label="Negatif", command=Im.negatif)
Couleur.add_command(label="Niveau de gris", command=Im.niveauGris)
Couleur.add_cascade(label="Changement de palette", menu=Change_palette)
Change_palette.add_command(label="Automne", command=Im.autumn)
Change_palette.add_command(label="Arc en ciel", command=Im.rainbow)
Change_palette.add_command(label="Bone", command=Im.bone)
Change_palette.add_command(label="Cool", command=Im.cool)
Change_palette.add_command(label="Hot", command=Im.hot)
Change_palette.add_command(label="HSV", command=Im.hsv)
Change_palette.add_command(label="Jet", command=Im.jet)
Change_palette.add_command(label="Ocean", command=Im.ocean)
Change_palette.add_command(label="Pink", command=Im.pink)
Change_palette.add_command(label="Printemps", command=Im.spring)
Change_palette.add_command(label="Ete", command=Im.summer)
Change_palette.add_command(label="Hiver", command=Im.winter)
## Ajout d'un onglet Texture
menubar.add_cascade(label="Texture", menu=Texture)
Texture.add_command(label="Camouflage", command=Im.camo)
## Ajout d'un onglet Contours
menubar.add_cascade(label="Contours", menu=Contour)
Contour.add_command(label="Extraction", command=Im.extraction)
Contour.add_command(label="Detection", command=Im.detection)
## Ajout d'un onglet A propos de
menubar.add_cascade(label="?", menu=Aide)
Aide.add_command(label="A propos de ...", command=exit)
trait.transient(root)
trait.mainloop()
