Adaptive image filtering based on a distance transform
View Patent ↗A method for reducing visual artifacts in reconstructed images. One embodiment of the method determines edge energy for each pixel in the image and then compares the edge energy for each pixel to a threshold, producing an edge map. A distance transform is then used to produce a filter map and a filter is applied to pixel in the image, such that the filter applied is dependent upon a filter map value for each pixel. An output value for each pixel is then produced.
1. A method for reducing visual artifacts in reconstructed images, the method comprising:
determining edge energy for each pixel in the image;
comparing the edge energy for each pixel to a threshold, producing an edge map, wherein the threshold is selected to be the minimum of a user-defined maximum threshold and half of the maximum edge energy within a current image block;
using a distance transform to produce a filter map;
applying a filter to each pixel in the image, wherein the filter applied is dependent upon a filter map value for each pixel; and
producing an output value for each pixel in the image.
2. The method of claim 1 , wherein the edge energy is determined using a Sobel filter.
3. The method of claim 1 , wherein the edge energy is determined using a Prewitt filter.
4. The method of claim 1 , wherein the edge energy is determined using a derivative of Gaussian filter.
5. The method of claim 1 , wherein the edge energy is determined using a normal filter.
6. The method of claim 1 , wherein pixels with an edge energy above the threshold are labeled as edge pixels and the corresponding value set to 1 in the edge map.
7. The method of claim 1 wherein the user-defined maximum threshold is 184.
8. The method of claim 1 , wherein boundary pixels in the edge map are set as edge pixels.
9. The method of claim 1 , wherein using a distance transform, applying a filter to each pixel in the image, and producing an output value for each pixel in the image of are not performed if the threshold is less than a predefined minimum value.
10. The method of claim 1 , wherein the distance transform is the minimum value of a user-defined maximum distance and a distance of a current pixel to a nearest edge in the edge map.
11. The method of claim 10 , wherein the user-defined maximum distance is 3 pixels.
12. The method of claim 1 , wherein edge pixels in the edge map are given a filter map value of zero.
13. The method of claim 1 , wherein the filter further comprises a two-dimensional low pass filter.
14. The method of claim 1 , wherein the filter further comprises two one-dimensional low pass filters.
15. The method of claim 1 , wherein the filter map further comprises separate filter maps, one for horizontal operators and one for vertical operators.
16. The method of claim 1 , wherein the reconstructed image is in RGB color space.
17. The method of claim 1 , wherein the reconstructed image is in LAB color space.
18. A facsimile machine, comprising:
a receiver operable to receive an input image;
a processor operable to decompress the input image, producing a decompressed image;
a postprocessor operable to:
determine edge energy for each pixel in the image;
compare the edge energy for each pixel to a threshold, producing an edge map, wherein the threshold is selected to be the minimum of a user-defined maximum threshold and half of the maximum edge energy within a current image block;
use a distance transform to produce a filter map;
apply a filter to each pixel in the image, wherein the filter applied is dependent upon a filter map value for each pixel; and
produce an output value for each pixel in the image; and
print engine operable to print the decompressed and adaptively filtered image.
19. The method of claim 18 , wherein the input image is a color image.
20. The method of claim 18 , wherein the processor and the postprocessor are the same processor performing different sets of instructions.
21. An article of computer-readable media including instructions that, when executed, cause the computer to:
determine edge energy for each pixel in the image;
compare the edge energy for each pixel to a threshold, producing an edge map, wherein the threshold is selected to be the minimum of a user-defined maximum threshold and half of the maximum edge energy within a current image block;
use a distance transform to produce a filter map;
apply a filter to each pixel in the image, wherein the filter applied is dependent upon a filter map value for each pixel; and
produce an output value for each pixel in the image.