Encoding documents using pixel classification-based preprocessing and JPEG encoding
View Patent ↗A method for JPEG encoding an image that includes preprocessing steps performed prior to a JPEG compression step. The preprocessing steps include transforming block image data, using a variable quantization table selected for each block according to the classification of pixels in that block to produce quantized image data, and inverse transforming the quantized image data to produce filtered image data.
1. A method for encoding an image including a plurality of pixels, comprising:
classifying each pixel of the image as one of a plurality of pixel classifications;
defining the image as a plurality of blocks of pixels;
for each block, classifying that block as one of a plurality of block classifications based on the pixel classifications of pixels in that block;
for each block, performing a discrete cosine transform upon that block to produce a transformed matrix corresponding to that block and including transform coefficients for that block;
for each transformed matrix and corresponding block, quantizing the transform coefficients of that transformed matrix using a selected quantization table according to the block classification of that block to produce a quantized matrix corresponding to that block;
for each quantized matrix and corresponding block, performing an inverse discrete cosine transform on that quantized matrix to produce a filtered image portion, the filtered image portions together comprising a filtered image; and
performing a JPEG compression on the filtered image,
wherein a specific pixel is classified by analyzing the intensity values of a plurality of pixels in an N by N window centered around that specific pixel, the analysis including:
determining a first maximum value of a first plurality of values that includes the sums of the intensity values of each row of the N by N window, determining a first minimum value of that first plurality of values, and calculating a first difference between the first maximum value and the first minimum value;
determining a second maximum value of a second plurality of values that includes the sums of the intensity values of each column of the N by N window, determining a second minimum value of that second plurality of values, and calculating a second difference between the second maximum value and the second minimum value; and
classifying a pixel as halftone if the first difference and the second difference are less than or equal to a first predefined threshold T1, and
wherein said method is performed by a processor.
2. The method of claim 1 , wherein N=17.
3. The method of claim 1 , further including classifying a pixel as a potential text element if the first difference or the second difference is greater than the first predefined threshold T 1 .
4. The method of claim 1 , further including classifying a potential text element as text if a modified line detector kernel of size M by M centered around the pixel of interest indicates the presence of text.
5. The method of claim 1 , further including the steps of:
summing the intensity values along the two diagonals of the N by N window to produce a first sum and a second sum, and determining a third difference between the first sum and the second sum; and
classifying a pixel as halftone if the first difference and the second difference are less than or equal to a first predefined threshold T1 and the absolute value of the third difference is less than or equal to a second predefined threshold T2.
6. The method of claim 5 , further including classifying a pixel as a potential text element if the first difference or the second difference is greater than the first predefined threshold T1 or the third difference is greater than the second predefined threshold T2.
7. The method of claim 6 , further including classifying a potential text element as text if a modified line detector kernel of size M by M centered around the pixel of interest indicates the presence of text.
8. A method for encoding an image including a plurality of pixels, comprising:
classifying each pixel of the image as one of a plurality of pixel classifications;
defining the image as a plurality of blocks of pixels;
for each block, classifying the block as one of a plurality of block classifications based on the pixel classifications of pixels in that block, performing a discrete cosine transform on the block to produce a transformed block including transform coefficients, quantizing the transform coefficients using a selected quantization table according to the block classification to produce a quantized block, and performing an inverse discrete cosine transform on the quantized block to produce a filtered image portion;
combining the filtered image portions for all the blocks to produce a filtered image; and
performing a JPEG compression on the filtered image,
wherein a specific pixel is classified by analyzing the intensity values of a plurality of pixels in an N by N window centered around that specific pixel, the analysis including:
determining a first maximum value of a first plurality of values that includes the sums of the intensity values of each row of the N by N window, determining a first minimum value of that first plurality of values, and calculating a first difference between the first maximum value and the first minimum value;
determining a second maximum value of a second plurality of values that includes the sums of the intensity values of each column of the N by N window, determining a second minimum value of that second plurality of values, and calculating a second difference between the second maximum value and the second minimum value; and
classifying a pixel as halftone if the first difference and the second difference are less than or equal to a first predefined threshold T1, and
wherein said method is performed by a processor.
9. The method of claim 8 , wherein N=17.
10. The method of claim 8 , further including classifying a pixel as a potential text element if the first difference or the second difference is greater than the first predefined threshold T1.
11. The method of claim 10 , further including classifying a potential text element as text if a modified line detector kernel of size M by M centered around the pixel of interest indicates the presence of text.
12. The method of claim 8 , further including the steps of summing the intensity values along the two diagonals of the N by N window to produce a first sum and a second sum, determining a third difference between the first sum and the second sum, and only classifying a pixel as halftone if the absolute value of the third difference is less than or equal to a second predefined threshold T2.
13. The method of claim 12 , further including classifying a pixel as a potential text element if the first difference or the second difference is greater than the first predefined threshold T1.
14. The method of claim 13 , further including classifying a potential text element as text if a modified line detector kernel of size M by M centered around the pixel of interest indicates the presence of text.