datasets/raise/extract_noise.py

@@ -33,34 +33,35 @@ if requiresRaiseFiltering:
 
     imagesFileNames = [imageFileName for imageFileName in tqdm(imagesFileNames, 'Filtering images') if files[imageFileName]['Device'] == 'Nikon D7000' and Image.open(f'{imagesFolderPath}/{imageFileName}').size == (4946, 3278)]
 
-minGreen = 308#None
-
 def treatImage(imageFileName):
-    global mean, numberOfImagesInMean, minGreen
+    global mean, numberOfImagesInMean
     imageFilePath = f'{imagesFolderPath}/{imageFileName}'
     if imageFileName.endswith('.NEF'):
+        imageFileName = 'ra2c888f8t.NEF'
+        imageFilePath = f'{imagesFolderPath}/{imageFileName}'
+        print(imageFilePath)
         with rawpy.imread(imageFilePath) as raw:
             colorDesc = raw.color_desc.decode('ascii')
             assert colorDesc == 'RGBG'
-            assert np.array_equal(raw.raw_pattern, np.array([[0, 1], [3, 2]], dtype = np.uint8))
+            assert np.array_equal(raw.raw_pattern, np.array([[3, 2], [0, 1]], dtype = np.uint8))
             rawImageVisible = raw.raw_image_visible.copy()
-            greenRawImageVisible = rawImageVisible[1::2, ::2]
-            '''
-            greenRawImageVisibleMin = greenRawImageVisible.min()
-            if minGreen is None or greenRawImageVisibleMin < minGreen:
-                minGreen = greenRawImageVisibleMin
-                print(minGreen)
-            '''
-            imageNpArray = (greenRawImageVisible - minGreen) / (4908 - minGreen)
+            centerY = 2263
+            centerX = 237
+            RANGE = 5
+            rawImageVisible = rawImageVisible.astype(np.float64)
+            rawImageVisibleShape = rawImageVisible.shape
+            for y in range(0, rawImageVisibleShape[0], 2):
+                for x in range(0, rawImageVisibleShape[1], 2):
+                    rawImageVisible[y, x] *= -1
+            print(rawImageVisible[centerY - RANGE:centerY + RANGE, centerX - RANGE:centerX + RANGE])
+            #greenRawImageVisible = raw.rawImageVisible
             # Pay attention to range of values expected by the denoiser.
-            # Indeed if provide the thousands valued raw image, then the denoiser only returns values between 0 and 1 and making the difference between both look pointless.
-            #exit(1)
+            exit(1)
     else:
         imagePil = Image.open(imageFilePath)
         imageNpArray = img_as_float(np.array(imagePil))
     # As the arguments differ from a denoiser to the other should use a match instead.
-    imageDenoisedNpArray = denoise(imageNpArray, rescale_sigma=True)
-    #print(imageNpArray, imageDenoisedNpArray)
+    imageDenoisedNpArray = denoise(imageNpArray, channel_axis=-1, convert2ycbcr=True, rescale_sigma=True)
     imageNoiseNpArray = imageNpArray - imageDenoisedNpArray
     if mean is None:
         mean = imageNoiseNpArray