Methods and apparatus for object recognition and compression

1. A method of recognizing and compressing an image for transmission such that a requirement for transmission of the image is reduced while maintaining target-specific utility of said image, comprising:

defining a first object class having a first object criteria that is at least partially related to the target-specific utility of said image;

recognizing an object within said image as a member of said first object class if said object substantially meets said first object criteria of said first object class;

decomposing said image into a plurality of subbands;

providing to each of said multiple subbands a mapping of the region of said image having said object recognized as a member of said first object class;

creating for each of said multiple subbands a first subband class sequence corresponding to said region of said image having said object recognized as a member of said first object class and a second subband class corresponding to regions of said image not having said object recognized as a member of said first object class;

normalizing said first and said second subband class sequences; and

encoding said normalized first subband class sequences at a first coding rate and encoding said normalized second subband class sequences at a second coding rate, wherein said first coding rate is higher than said second coding rate.

2. The method of claim 1 , further comprising synthesizing an object contour of said object within the image.

3. The method of claim 1 , further comprising synthesizing a rotated binary image chip of said object within the image.

4. The method of claim 1 , further comprising synthesizing a symmetrically rotated binary image chip of said object within the image.

5. The method of claim 2 , wherein synthesizing said object contour of said object within the image comprises:

producing a binary image of the image; and

extracting an outer edge of the binary image.

6. The method of claim 2 , wherein synthesizing said object contour of said object within the image comprises:

producing an edge image of the image; and

extracting an outer edge of the edge image.

7. The method of claim 5 , further comprising:

generating a coordinate list of said outer edge of said binary image, said coordinate list specifying a bounding region enclosing said object within the image;

extracting an image chip from the image corresponding to said bounding region specified by said coordinate list;

generating a binary image chip of said image chip; and

conforming said binary image chip to a symmetrical axis to product said object contour.

8. The method of claim 6 , further comprising:

generating a coordinate list with said outer edge of said edge image, said coordinate list specifying a bounding region enclosing said object within said image;

extracting an image chip from the image corresponding to said bounding region specified by said coordinate list;

generating a binary image chip of said image chip; and

conforming said binary image chip to a symmetrical axis to produce said object contour.

9. The method of claim 7 , wherein generating said coordinate list of said outer edge includes:

comparing the dimension of said bounding region to predetermined validation dimension; and

validating said object if the dimension of said bounding region is less than said predetermined validation dimension.

10. The method of claim 8 , wherein generating a coordinate list of said outer edge includes:

comparing the dimensions of said bounding region to predetermined validation dimensions; and

validating said object if the dimensions of said bounding region is less than said predetermined validation dimensions.

11. The method of claim 7 , wherein said symmetrical axis is a vertically symmetrical axis.

12. The method of claim 8 , wherein said symmetrical axis is a vertically symmetrical axis.

13. The method of claim 1 , wherein recognizing said object within the image further comprises recognizing said object within the image as a member of a first object sub-class of said first object class if said object substantially meets said first object criteria of said first object class and said first sub-class object criteria of said first object sub-class.

14. An apparatus for recognizing and compressing an image such that a requirement for the transmission of said image is reduced while maintaining target-specific utility of said image, comprising:

an object processor configured to receive the image and synthesize an object existing within said image;

a classification engine configured to recognize said object existing within said image as a member of a first object class if said object substantially meets first object criteria of said first object class that is at least partially related to the target-specific utility of said image;

an image decomposer configured to decompose said image into a plurality of subbands;

a wavelet mask constructor configured to provide for each of said multiple subbands a mapping of the region of said image having said object;

a subband class sequence generator configured to create for each of said multiple subbands a first subband class sequence corresponding to said region of said image having said object recognized as a member of said first object class and a second subband class corresponding to regions of said image not having said object recognized as a member of said first object class;

a normalizer configured to normalize said first and said second subband class sequences; and

a multi-rate encoder configured to encode said normalized first subband class sequences at a first coding rate and to encode said normalized second subband class sequences at a second coding rate, wherein said first coding rate is higher than said second coding rate.

15. The apparatus of claim 14 , wherein said object is synthesized to form a rotated binary image chip of said object existing within the image.

16. The apparatus of claim 14 , wherein said object is synthesized to form a symmetrically rotated binary image chip of said object existing within the image.

17. The apparatus of claim 14 , wherein said object processor comprises:

a binary image generator configured to receive the image and produce a binary image of said object; and

an outer edge extractor configured to extract an outer edge of the binary image and generate a coordinate list with said outer edge that specifies a bounding region enclosing said object existing within the image.

18. The apparatus of claim 14 , wherein said object processor comprises:

an edge image generator configured to receive the image and produce an edge image of said object; and

an outer edge extractor configured to extract an outer edge of the edge image and generate a coordinate list with said outer edge that specifies a bounding region enclosing said object existing within the image.

19. The apparatus of claim 17 , further comprising:

an object extractor configured to extract an image chip from the image corresponding to said bounding region specified by said coordinate list;

a binary chip image generator configured to generate a binary image chip of said image chip; and

a symmetrical rotator configured to conform said binary image chip to a symmetrical axis to produce said object.

20. The apparatus of claim 18 , further comprising:

an object extractor configured to extract an image chip from the image corresponding to said bounding region specified by said coordinate list;

a binary chip image generator configured to generate a binary image chip of said image chip; and

a symmetrical rotator configured to conform to said binary image chip to a symmetrical axis to produce said object.

21. The apparatus of claim 1 wherein the size of said region having said object recognized as a member of said first object class in said mapping of said region having said object recognized as a member of said first object class within each of said multiple subbands is adjusted to correspond to the size in said image of said object recognized as a member of said first object class.

22. The apparatus of claim 14 wherein said wavelet mask constructor adjusts the size of said region having said object recognized as a member of said first object class in said mapping of said region having said object recognized as a member of said first object class within each of said multiple subbands is adjusted to correspond to the size in said image of said object recognized as a member of said first object class.