JPEG encoding for document images using pixel classification
View Patent ↗The invention provides a system and method for JPEG encoding an image that includes variable quantization for each block, depending upon the classification of each block, which is dependent on the classification of pixels in that block.
1. A system for encoding an image comprising:
a pixel classification module for classifying each pixel of the image as one of a plurality of pixel classifications;
a transform calculator for defining the image as a plurality of blocks of pixels and transforming each block to produce transform coefficients for the block;
a block classification module for classifying each block as one of a plurality of block classifications based on the pixel classifications of pixels in the block;
a quantization table selector unit operating to select, for each block, a quantization table according to the block classification for that block; and
a quantizer for quantizing the transformed coefficients for each block using the selected quantization table for that block.
2. The system of claim 1 , wherein the transform calculator uses discrete cosine transforms.
3. The system of claim 1 , wherein a JPEG-compliant encoding is performed.
4. The system of claim 1 , wherein the pixel classification module classifies each pixel as one of text and picture.
5. The system of claim 1 , wherein the pixel classification module classifies each pixel as one of text, picture and background.
6. The system of claim 1 , wherein the pixel classification module computes a vector having elements representing comparisons between neighboring pixels.
7. The system of claim 6 , wherein the vector includes elements representing comparisons of pixel values between adjacent pixel locations.
8. The system of claim 6 , wherein the pixel classification module computes a norm for the vector and classifies the pixel based on the value of the norm.
9. The system of claim 6 , wherein the pixel classification module computes a norm for the vector, defines the pixel as a potential text element if the value of the norm is above a first threshold, and classifies the potential text element as text if the distance between it and the closest potential text element is less than a second threshold.
10. The system of claim 6 , wherein a first element of the vector corresponds to a difference in gray values of pixels horizontally adjacent to the given pixel and a second element of the vector corresponds to a difference in gray values of pixels vertically adjacent to the given pixel.
11. The system of claim 1 , wherein the plurality of block classifications includes text and picture.
12. The system of claim 1 , wherein the plurality of block classifications includes text, picture and mixed.
13. A method for encoding an image comprising the steps of:
classifying each pixel of the image as one of a plurality of pixel classifications;
defining the image as a plurality of blocks of pixels;
transforming each block to produce transform coefficients for the block;
classifying each block as one of a plurality of block classifications based on the pixel classifications of pixels in the block;
selecting a quantization table for each block according to its block classification; and
quantizing the transform coefficients for each block using the selected quantization table for that block.
14. The method of claim 13 , wherein the plurality of pixel classifications include text and picture.
15. The method of claim 13 , wherein the plurality of pixel classifications include text, picture and background.
16. The method of claim 13 , wherein the step of classifying each pixel includes computing a vector having elements representing comparisons between neighboring pixels.
17. The method of claim 16 , wherein the step of classifying each pixel includes computing a vector having elements representing comparisons of gray values between adjacent pixels.
18. The method of claim 16 , wherein the step of classifying each pixel further includes computing a norm for the vector, and classifying the pixel based on the value of the norm.
19. The method of claim 16 , wherein the step of classifying each pixel further includes computing a norm for the vector, defining the pixel as a potential text element if the value of the norm is above a first threshold, and classifying the pixel as text if the distance between it and the closest potential text element is less than a second threshold.
20. The method of claim 16 , wherein a first element of the vector corresponds to a difference in the gray values of pixels horizontally adjacent to the given pixel and a second element of the vector corresponds to a difference in the gray values of pixels vertically adjacent to the given pixel.
21. The method of claim 13 , wherein the plurality of block classifications includes text and picture.
22. The method of claim 13 , wherein the plurality of block classifications includes text, picture and mixed.