System and method for bad pixel replacement in image processing
View Patent ↗A system and method is provided for processing a digital image. The system and method processes image data by replacing bad pixel data in the digital image. Specifically, the system and method replaces bad pixel data in the image data by comparing each pixel to selected neighboring pixels, including pixels of other colors in some cases, and determining if the pixel is likely corrupt, such as from a detective pixel. Corrupted pixels are then replaced using averages, means, maximums, or other statistical functions of select nearby pixels.
1. A method of processing a digital image in a processor tangibly embodying a program of instructions to perform method steps, the method comprising the steps of:
receiving image data defining the image, the image data including a plurality of pixels, the plurality of pixels including a plurality of green pixels, a plurality of red pixels, and a plurality of blue pixels;
determining for each pixel if the pixel is likely corrupt by comparing each pixel to a plurality of neighboring pixels, wherein the step of determining for each pixel if the pixel is likely corrupt by comparing each pixel to a plurality of neighboring pixels comprises:
for each of the plurality of green pixels:
determining if the green pixel exceeds a plurality of green-neighboring pixels by a first green threshold difference;
determining if the green pixel is less than the plurality of green-neighboring pixels by a second green threshold difference;
for each of the plurality of red pixels:
determining if the red pixel exceeds a plurality of red-neighboring pixels by a first red threshold difference;
determining if the red pixel is less than the plurality of red-neighboring pixels by a second red threshold difference;
determining if an area around the red pixel is locally flat by determining a maximum difference in neighboring green pixels and comparing the maximum difference in neighboring green pixels to a first local flatness threshold;
determining if the area of the red pixel is locally dark by determining a brightness in neighboring green pixels and comparing the brightness to a first brightness threshold; and
for each of the plurality of blue pixels:
determining if the blue pixel exceeds a pair of a plurality of blue-neighboring pixels by a first blue threshold difference;
determining if the blue pixel is less than a pair of the plurality blue-neighboring pixels by a second blue threshold difference;
determining if an area around the blue pixel is locally flat by determining a maximum difference in neighboring green pixels and comparing the maximum difference in neighboring green pixels to a second local flatness threshold;
determining if the area of the blue pixel is locally dark by determining a brightness in neighboring green pixels and comparing the brightness to a second brightness threshold; and
replacing each pixel determined to be likely corrupt with a statistical function of a plurality of nearby pixels.
2. The method of claim 1 wherein the step of replacing each pixel determined to be likely corrupt with a statistical function of a plurality of nearby pixels comprises replacing the pixel with an average of the plurality of nearby pixels.
3. The method of claim 1 wherein the step of replacing each pixel determined to be likely corrupt with a statistical function of a plurality of nearby pixels comprises replacing the pixel with a maximum of the plurality of nearby pixels.
4. The method of claim 1 wherein the step of replacing each pixel determined to be likely corrupt with a statistical function of a plurality of nearby pixels comprises replacing the pixel with a mean of the plurality of nearby pixels.
5. The method of claim 1
wherein the step of replacing each pixel determined to be likely corrupt with a statistical function of a plurality of nearby pixels comprises replacing each green pixel determined to be likely corrupt with an average of a plurality of nearby pixels.
6. The method of claim 5 wherein the plurality of green-neighboring pixels comprise four closest green pixels and wherein the step of replacing each pixel determined to be likely corrupt with a statistical function of a plurality of nearby pixels comprises replacing green pixels determined to be likely corrupt with an average of the four closest green pixels.
7. The method of claim 1
wherein the step of replacing each pixel determined to be likely corrupt with a statistical function of a plurality of nearby pixels comprises replacing each red pixel determined to be likely corrupt with a statistical function of a plurality of nearby pixels.
8. The method of claim 7 wherein the plurality of red-neighboring pixels comprise two green pixels and two red pixels and wherein the step of replacing each pixel determined to be likely corrupt with a statistical function of a plurality of nearby pixels comprises replacing red pixels determined to be likely corrupt with a mean of two green pixels and two red pixels.
9. The method of claim 7 wherein the statistical function of a plurality of nearby pixels comprises a maximum of the plurality of nearby pixels if the red pixel exceeds the plurality of neighboring pixels by the first red threshold difference, and wherein the statistical function of a plurality of nearby pixels comprises a maximum of the plurality of nearby pixels if the red pixel exceeds the plurality of neighboring pixels by the first red threshold difference and if the red pixel is in a locally flat area or a locally dark area, and wherein the statistical function of a plurality of nearby pixels comprises a mean the plurality of nearby pixels if the red pixel is less than the plurality neighboring pixels by the second red threshold difference, and wherein the statistical function of a plurality of nearby pixels comprises a mean of the plurality of nearby pixels if the red pixel is less than the plurality neighboring pixels by the second red threshold difference and if the red pixel is in a locally flat area.
10. The method of claim 1
wherein the step of replacing each pixel determined to be likely corrupt with a statistical function of a plurality of nearby pixels comprises replacing each blue pixel determined to be likely corrupt with a statistical function of a plurality of nearby pixels.
11. The method of claim 10 wherein the plurality of blue-neighboring pixels comprise four blue pixels, and wherein the step of replacing each pixel determined to be likely corrupt with a statistical function of a plurality of nearby pixels comprises replacing blue pixels determined to be likely corrupt with a mean of four blue pixels.
12. The method of claim 10 wherein the statistical function of a plurality of nearby pixels comprises a maximum of the plurality of nearby pixels if the blue pixel exceeds the plurality of neighboring pixels by the first blue threshold difference and if the blue pixel is in a locally flat area or a locally dark area, and wherein the statistical function of a plurality of nearby pixels comprises a mean of the plurality of nearby pixels if the blue pixel is less than the plurality neighboring pixels by the second blue threshold difference and if the blue pixel is in a locally flat area, and wherein the statistical function of a plurality of nearby pixels comprises a mean of the plurality of nearby pixels if the blue pixel is less than the plurality neighboring pixels by the second blue threshold difference and if the blue pixel is in a locally flat area and if the blue pixel is in a locally dark area.
13. A method of processing an a digital image in a digital camera, the method comprising the steps of:
generating image data defining the image with the digital camera, the image data including a plurality of pixels;
for each green pixel in the plurality of pixels, replacing likely corrupt green pixels by:
determining if a green pixel exceeds first selected green-neighboring pixels by a first green threshold difference and replacing the green pixel with an average of first selected green-nearby pixels if the green pixel exceeds the first selected green-neighboring pixels by the first green threshold difference; and
determining if the green pixel is less than second selected green-neighboring pixels by a second green threshold difference and replacing the green pixel with an average of selected green-nearby pixels if the green pixel is less than the second selected green-neighboring pixels by the second green threshold difference; and
for each red pixel in the plurality of pixels, replacing likely corrupt red pixels by:
determining if a red pixel exceeds first selected red-neighboring pixels by a first red threshold difference and replacing the red pixel with a maximum of first selected red-nearby pixels if the red pixel exceeds the first selected red-neighboring pixels by the first red threshold difference; and
determining if the red pixel exceeds second selected red-neighboring pixels by a second red threshold difference and determining if the red pixel is in a locally flat area and determining if the red pixel is in a locally dark area and replacing the red pixel with a maximum of second selected red-nearby pixels if the red pixel exceeds the second selected red-neighboring pixels by the second red threshold difference and if the red pixel is in a locally flat area or a locally dark area; and
determining if the red pixel is less than third selected red-neighboring pixels by a third red threshold difference and replacing the red pixel with a mean of third selected red-nearby pixels if the red pixel is less than the third selected red-neighboring pixels by the third red threshold difference; and
determining if the red pixel is less than fourth selected red-neighboring pixels by a fourth red threshold difference and determining if the red pixel is in a locally flat area and replacing the red pixel with a mean of fourth selected red-nearby pixels if the red pixel is less than the fourth selected red-neighboring pixels by the fourth red threshold difference and if the red pixel is in a locally flat area; and
for each blue pixel in the plurality of pixels, replacing likely corrupt blue pixels by:
determining if a blue pixel exceeds any pair of first selected blue-neighboring pixels by a first blue threshold difference determining if the blue pixel is in a locally flat area and determining if the blue pixel is in a locally dark area and replacing the blue pixel with a mean of first selected blue-nearby pixels if the blue pixel exceeds the first selected blue-neighboring pixels by the first blue threshold difference and if the blue pixel is in a locally flat area or a locally dark area; and
determining if the blue pixel is less than second selected blue-neighboring pixels by a second blue threshold difference and determining if the blue pixel is in a locally flat area and replacing the blue pixel with a mean of second selected blue-nearby pixels if the blue pixel is less than the second selected blue-neighboring pixels by the second blue threshold difference and if the blue pixel is in a locally flat area; and
determining if the pixel is less than third selected blue-neighboring pixels by a third blue threshold difference and determining if the blue pixel is in a locally flat area and determining if the blue pixel is in a locally dark area and replacing the blue pixel with a mean of third selected blue-nearby pixels if the blue pixel is less than the third selected blue-neighboring pixels by the third blue threshold difference if the blue pixel is in a locally flat area and in a locally dark area.
14. The method of claim 13 wherein the plurality of red-nearby pixels comprise two green pixels and two red pixels.
15. A processor, tangibly embodying a program of instructions to perform method steps for digital image processing, comprising the machine executed steps of:
receiving image data defining the image, the image data including a plurality of pixels;
for each green pixel in the plurality of pixels, replacing likely corrupt green pixels by:
determining if a green pixel exceeds first selected green-neighboring pixels by a first green threshold difference and replacing the green pixel with an average of first selected green-nearby pixels if the green pixel exceeds the first selected green-neighboring pixels by the first green threshold difference; and
determining if the green pixel is less than second selected green-neighboring pixels by a second green threshold difference and replacing the green pixel with an average of selected green-nearby pixels if the green pixel is less than the second selected green-neighboring pixels by the second green threshold difference; and
for each red pixel in the plurality of pixels, replacing likely corrupt red pixels by:
determining if a red pixel exceeds first selected red-neighboring pixels by a first red threshold difference and replacing the red pixel with a maximum of first selected red-nearby pixels if the red pixel exceeds the first selected red-neighboring pixels by the first red threshold difference; and
determining if the red pixel exceeds second selected red-neighboring pixels by a second red threshold difference and determining if the red pixel is in a locally flat area and determining if the red pixel is in a locally dark area and replacing the red pixel with a maximum of second selected red-nearby pixels if the red pixel exceeds the second selected red-neighboring pixels by the second red threshold difference and if the red pixel is in a locally flat area or a locally dark area; and
determining if the red pixel is less than third selected red-neighboring pixels by a third red threshold difference and replacing the red pixel with a mean of third selected red-nearby pixels if the red pixel is less than the third selected red-neighboring pixels by the third red threshold difference; and
determining if the red pixel is less than fourth selected red-neighboring pixels by a fourth red threshold difference and determining if the red pixel is in a locally flat area and replacing the red pixel with a mean of fourth selected red-nearby pixels if the red pixel is less than the fourth selected red-neighboring pixels by the fourth red threshold difference and if the red pixel is in a locally flat area; and
for each blue pixel in the plurality of pixels, replacing likely corrupt blue pixels by:
determining if a blue pixel exceeds any pair of first selected blue-neighboring pixels by a first blue threshold difference determining if the blue pixel is in a locally flat area and determining if the blue pixel is in a locally dark area and replacing the blue pixel with a mean of first selected blue-nearby pixels if the blue pixel exceeds the first selected blue-neighboring pixels by the first blue threshold difference and if the blue pixel is in a locally flat area or a locally dark area; and
determining if the blue pixel is less than second selected blue-neighboring pixels by a second blue threshold difference and determining if the blue pixel is in a locally flat area and replacing the blue pixel with a mean of second selected blue-nearby pixels if the blue pixel is less than the second selected blue-neighboring pixels by the second blue threshold difference and if the blue pixel is in a locally flat area; and
determining if the pixel is less than third selected blue-neighboring pixels by a third blue threshold difference and determining if the blue pixel is in a locally flat area and determining if the blue pixel is in a locally dark area and replacing the blue pixel with a mean of third selected blue-nearby pixels if the blue pixel is less than the third selected blue-neighboring pixels by the third blue threshold difference if the blue pixel is in a locally flat area and in a locally dark area.
16. The processor of claim 15 wherein the plurality of red-nearby pixels comprise two green pixels and two red pixels.