First PRNU distance test

algorithms/context_adaptive_interpolator/context_adaptive_interpolator.py

@@ -6,14 +6,15 @@ from wiener_filter import wienerFilter
 
 # Assume greyscale PIL image passed.
 # What about other color channels? See #11.
-def contextAdaptiveInterpolator(I, IImage):
+def contextAdaptiveInterpolator(I, IImage, showProgress = False):
     rImage = Image.new('L', (IImage.size[0] - 2, IImage.size[1] - 2))
     r = rImage.load()
 
     # This threshold is debatable. See #13.
     THRESHOLD = 20
 
-    print('before for loops')
+    if showProgress:
+        print('before for loops')
     # Equation (10)
     # Accelerate computation. See #15.
     for m in range(1, IImage.size[0] - 1):
@@ -40,7 +41,8 @@ def contextAdaptiveInterpolator(I, IImage):
                 else:
                     newPixel = I[m, n] - median(A)
                 r[m - 1, n - 1] = round(newPixel)
-    print('after for loops')
+    if showProgress:
+        print('after for loops')
 
     # Why need to rotate the image? See #14.
     #rImage.rotate(-90).show()
@@ -49,5 +51,6 @@ def contextAdaptiveInterpolator(I, IImage):
     # $\sigma_0^2$ is the noise variance.
     sigma_0 = 9 ** 0.5
 
-    print('before wiener filter')
-    return wienerFilter(r, rImage, Q, sigma_0)
+    if showProgress:
+        print('before wiener filter')
+    return wienerFilter(r, rImage, Q, sigma_0, showProgress)

algorithms/context_adaptive_interpolator/wiener_filter.py

@@ -1,7 +1,7 @@
 from PIL import Image
 from tqdm import tqdm
 
-def wienerFilter(r, rImage, Q, sigma_0):
+def wienerFilter(r, rImage, Q, sigma_0, showProgress):
     h_wImage = Image.new('L', (rImage.size[0], rImage.size[1]))
     h_wImagePixels = h_wImage.load()
 
@@ -29,10 +29,13 @@ def wienerFilter(r, rImage, Q, sigma_0):
         B_q = [(x, z) for x in getPixelIndexesAround(i, numberOfPixelsInEachDirection) for z in getPixelIndexesAround(j, numberOfPixelsInEachDirection)]
         return max(0, (1 / q ** 2) * sum([h[getPixelWithinImage(x, hImage.size[0]), getPixelWithinImage(z, hImage.size[1])] ** 2 - sigma_0 ** 2 for (x, z) in B_q]))
 
-    print('wiener filter start for loops')
-    for i in tqdm(range(rImage.size[0])):
+    if showProgress:
+        print('wiener filter start for loops')
+    rImageSize0Range = range(rImage.size[0])
+    for i in tqdm(rImageSize0Range) if showProgress else rImageSize0Range:
         for j in range(rImage.size[1]):
             h_wImagePixels[i, j] = round(h_w(rImage, r, i, j))
-    print('wiener filter end for loops')
+    if showProgress:
+        print('wiener filter end for loops')
 
     return h_wImage.rotate(-90)

datasets/fake/generate_dataset.py

@@ -7,11 +7,14 @@ sys.path.insert(0, '../../algorithms/distance/')
 
 from rmsdiff import rmsdiff
 
+sys.path.insert(0, '../../algorithms/context_adaptive_interpolator/')
+
+from context_adaptive_interpolator import contextAdaptiveInterpolator
+
 IMAGE_SIZE = 64
 
 def randomImage(scale):
-    # Is `np.clip` necessary? See `toPilImage`.
-    return np.random.normal(loc = 0, scale = scale, size = (IMAGE_SIZE, IMAGE_SIZE))
+    return np.maximum(np.random.normal(loc = 0, scale = scale, size = (IMAGE_SIZE, IMAGE_SIZE)), 0)
 
 prnu = randomImage(scale = 1)
 
@@ -19,12 +22,16 @@ images = [randomImage(scale = 10) + prnu for _ in range(10)]
 allImages = [prnu] + images
 
 def toPilImage(npArray):
-    nonNegativeArray = npArray - np.min(allImages)
-    nonNegativeArray = np.round(255 * nonNegativeArray / np.max(allImages))
-    return Image.fromarray(np.uint8(nonNegativeArray))
+    npArray = np.round(255 * npArray / np.max(allImages))
+    return Image.fromarray(np.uint8(npArray))
 
-toPilImage(prnu).show()
+prnu = toPilImage(prnu)
+#prnu.show()
+
+images = [toPilImage(image) for image in images]
+#images[0].show()
 
 for image in images:
-    print(rmsdiff(image, prnu))
-    print(rmsdiff(contextAdaptiveInterpolator(image), prnu))
+    initialRmsDiff = rmsdiff(image, prnu)
+    caiRmsDiff = rmsdiff(contextAdaptiveInterpolator(image.load(), image), prnu)
+    print(f'{initialRmsDiff=} {caiRmsDiff=}')