Method and apparatus for identifying regions of different content in an image

1. A method of identifying regions of different content in an image comprising:

using a processing unit to:

divide image data into a plurality of pixel blocks;

extract features of said pixel blocks; and

classify the content of said pixel blocks based on extracted features; and

wherein said extracting comprises for each pixel block:

convolving a magic square filter with the pixels of the pixel block and summing the results;

calculating the percentage of background pixels in the pixel block;

calculating the edge density of the pixel block; and

calculating the average saturation of the pixel block; and

wherein said classifying comprises:

initially classifying the content of each pixel block based on the magic square filter convolving and summing result and the percentage of background pixels; and

in the event that the content of a pixel block cannot be classified to a desired level of confidence, subsequently classifying the content of the pixel block based on at least one of the calculated edge density and average saturation.

2. The method of claim 1 wherein the extracted features are based on pixel block statistical information.

3. The method of claim 2 wherein said classifying is performed in cascading stages, later stages being employed only when earlier stages are unable to classify the pixel blocks.

4. The method of claim 3 wherein during said classifying, pixel blocks are identified as containing text content or non-text content.

5. The method of claim 4 further comprising:

aggregating connected pixel blocks identified as containing the same content type.

6. The method of claim 4 further comprising:

adjusting the borders of pixel blocks identified as containing text content to inhibit text from being truncated.

7. The method of claim 1 wherein said extracting further comprises for each pixel block:

calculating the number of uniform rectangles in the pixel block.

8. The method of claim 7 wherein, in the event that the pixel block cannot be classified during subsequent classifying, said classifying comprises further subsequently classifying the content of the pixel block based on the calculated edge density and the number of uniform rectangles.

9. The method of claim 8 wherein the average saturation, edge density and number of uniform rectangles are calculated only when subsequent classifying of the pixel block is required.

10. The method of claim 9 wherein during said classifying, pixel blocks are identified as containing text content or non-text content.

11. The method of claim 7 wherein the average saturation and edge density are calculated only when subsequent classifying of the pixel block is required.

12. The method of claim 1 wherein said image data is processed in bands.

13. The method of claim 12 further comprising receiving said bands of image data in a stream from an image scanning device.

14. The method of claim 1 wherein said classifying classifies pixel blocks as containing text content or non-text content.

15. An apparatus for identifying regions of different content in an image comprising:

a processing unit that comprises:

a feature extractor dividing image data into a plurality of pixel blocks and extracting features of said pixel blocks; and

a classifier classifying the content of said pixel blocks based on extracted features; and

wherein said feature extractor determines the results of a magic square filter convolved with each pixel block and at least one of background pixel percentage, edge density, number of uniform rectangles and average saturation thereby to extract said features; and

wherein said classifier initially classifies the content of each pixel block based on the magic square filter convolving results and the background pixel percentage; and in the event that the content of a pixel block cannot be classified to a desired level of confidence, said classifier subsequently classifies the content of the pixel block based on at least one of the calculated edge density and average saturation.

16. An apparatus according to claim 15 wherein said classifier comprises a plurality of classifier stages.

17. An apparatus according to claim 16 wherein said classifier classifies pixel blocks as containing text content or non-text content.

18. An apparatus according to claim 15 wherein said apparatus is at least part of a device selected from the group consisting of a photocopier, a facsimile machine and an all-in-one printer.

19. A non-transitory computer readable medium embodying a computer program for identifying regions of different content in an image, said computer program comprising:

computer program code for dividing image data into a plurality of pixel blocks;

computer program code for extracting features of said pixel blocks; and

computer program code for classifying the content of said pixel blocks based on extracted, features; and

wherein said computer program code for extracting comprises for each pixel block:

computer program code for convolving a magic square filter with the pixels of the pixel block and summing the results;

computer program code for calculating the percentage of background pixels in the pixel block;

computer program code for calculating the edge density of the pixel block; and

computer program code for calculating the average saturation of the pixel block; and

wherein said computer program code for classifying comprises:

computer program code for initially classifying the content of each pixel block based on the magic square filter convolving and summing result and the percentage of background pixels; and

computer program code for, in the event that the content of a pixel block cannot be classified to a desired level of confidence, subsequently classifying the content of the pixel block based on at least one of the calculated edge density and average saturation.