Benjamin_Loison/Robust_image_source_identification_on_modern_smartphones
1
0
Fork 0
Code Issues 68 Pull Requests Packages Projects Releases Wiki Activity

CLean remove_period_patterns.py and add debugging as unable to find back process to get wanted result

Browse Source
This commit is contained in:
Benjamin Loison 2024-05-23 01:55:25 +02:00
parent 0f5b9dcdaf
commit 81a994ed65
Signed by: Benjamin_Loison
SSH Key Fingerprint: SHA256:BtnEgYTlHdOg1u+RmYcDE0mnfz1rhv5dSbQ2gyxW8B8
1 changed files with 18 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
datasets/raise/fft/remove_period_patterns.py
View File

@ -10,7 +10,7 @@ sys.path.insert(0, '../')
from utils import Color, mergeSingleColorChannelImagesAccordingToBayerFilter
import matplotlib.pyplot as plt
PREFIX = 'mean_rafael_230424_mean_'
PREFIX = 'means/mean_rafael_230424_mean_'
X_STDDEV = 2
def getImageByColor(color):
@ -45,23 +45,20 @@ def removePeriodicPatterns(fft1Part):
kernel = Gaussian2DKernel(x_stddev = X_STDDEV)
# create a "fixed" image with NaNs replaced by interpolated values
fixedImage = interpolate_replace_nans(fft1Part, kernel)
return fixedImage
fixedImagePart = interpolate_replace_nans(fft1Part, kernel)
return fixedImagePart
'''
# TODO: are `.copy()` really necessary?
realFixedImage = removePeriodicPatterns(np.real(fft1).copy())
imaginaryFixedImage = removePeriodicPatterns(np.imag(fft1).copy())
fixedImage = realFixedImage + 1j * imaginaryFixedImage
'''
fixedImage = removePeriodicPatterns(abs(fft1))
fixedImage = imaginaryFixedImage
figure, axes = plt.subplots(1, 2, sharex = True, sharey = True)
plt.suptitle('Attenuating FFT significant lines')
axes[0].set_title('Original FFT')
#originalFft1Abs = abs(originalFft1)
#minValue = np.min()
firstImage = np.log10(1 + abs(originalFft1))
firstImage = np.log10(1 + abs(np.real(originalFft1)))
secondImage = np.log10(1 + abs(fixedImage))
images = [firstImage, secondImage]
vMin = np.min(images)
@ -71,20 +68,26 @@ axes[1].set_title('FFT with significant lines attenuated')
axes[1].imshow(secondImage, vmin = vMin, vmax = vMax)
plt.tight_layout()
plt.show()
#plt.imsave('fft.png', np.log10(1 + fixedImage))
figure, axes = plt.subplots(1, 2, sharex = True, sharey = True)
figure, axes = plt.subplots(1, 3, sharex = True, sharey = True)
def inverseFft(fft):
ifft2 = np.real(fftpack.ifft2(fftpack.ifftshift(fft)))
#ifft2 = np.maximum(0, np.minimum(ifft2, 255)) #* 255# / 255
return ifft2
invOriginalFft1 = inverseFft(originalFft1)
invFixedImage = inverseFft(fixedImage)
images = [invOriginalFft1, invFixedImage]
vMin = np.min(images)
vMax = np.max(images)
plt.suptitle('Rafael 23/04/24 PRNU mean denoiser with periodic patterns attenuated')
axes[0].set_title('Original PRNU')
invOriginalFft1 = inverseFft(originalFft1)
axes[0].imshow(invOriginalFft1)
axes[0].imshow(invOriginalFft1, vmin = vMin, vmax = vMax)
axes[1].set_title('PRNU with periodic patterns attenuated')
axes[1].imshow(inverseFft(fixedImage), vmin = invOriginalFft1.min(), vmax = invOriginalFft1.max())
axes[1].imshow(invFixedImage, vmin = vMin, vmax = vMax)
axes[2].set_title('Difference between both left images')
differenceBetweenBothImages = invOriginalFft1 - invFixedImage
# `np.log10(1 + abs(differenceBetweenBothImages))` does not seem more interesting.
axes[2].imshow(differenceBetweenBothImages)
plt.tight_layout()
plt.show()