Method and system for image compression using image symmetry

• US 7,742,649 B2
• Filed: 06/21/2007
• Issued: 06/22/2010
• Est. Priority Date: 12/17/2002
• Status: Expired due to Term

• Current Assignee: Symbol Technologies, LLC (Zebra Technologies Corporation)
• Original Assignee: Symbol Technologies, Inc. (Zebra Technologies Corporation)
• Inventors: He, Duanfeng, Tabesh, Ali
• Primary Examiner(s): Ge; Yuzhen
• Application Number:
  US11/766,228
• Publication Number:
  US 20070248275A1
• Time in Patent Office: 1,097 Days
• Field of Search: None
• US Class Current: 382/248
• CPC Class Codes: G06T 9/005 Statistical coding, e.g. Huffman, run length coding
• CPC Class Codes: G06V 40/16 Human faces, e.g. facial parts, sketches or expressions

Abstract

An image compression system and method is described, which makes use of the symmetry found in faces and heads to perform the compression. The image is divided along the line of symmetry, and pairs of corresponding pixels on the two divided sides are determined. A weighted average and a weighted variance of the pixel values of the pairs is computed, and is used to encode the image. A transform such as the Karhunen-Loeve transform is used to compute the weighted averages and variances.

Claims

1. A computer-implemented method for reducing an image representing a substantially symmetric subject, comprising the steps of:

dividing the image into a first part and a second part, the second part being substantially symmetric to the first part;

determining pairs of symmetrically corresponding pixels, a member of the pair in the first part having an X pixel value and a member of the pair in the second part having a Y pixel value;

computing a weighted average value A of the pixel values of each one of the pairs;

computing a weighted variance value B of the pixel values of each one of the pairs;

representing each of the pairs with the weighted average and the weighted variance values;

generating reduced image data including at least a plurality of A values and corresponding B values the pairs;

reconstituting the image by computing the X pixel value and the Y pixel value from the weighted average value A and the weighted variance value B; and

compressing the reduced image data with an image compression algorithm, wherein the reconstituting step further includes the substep of removing visual artifacts from the reconstituted image, wherein the visual artifacts comprise a seam line between the first part and the second part, wherein the dividing, determining, computing, representing and generating steps are executed by a microprocessor.

2. The method according to claim 1, further comprising the step of:

reconstituting the image by computing the X pixel value and the Y pixel value from the weighted average value A and the weighted variance value B using a scanner.

3. The method according to claim 2, further comprising the step of:

reconstituting the image as a function of an inverse transformation matrix L so that

X=I11A+I12B

Y=I21A+I22B

wherein I11, I12, I21, and I22 represent the elements of the L transform, based on statistical properties of the image.

4. The method according to claim 3, wherein the transformation matrix L is an inverse Karhunen-Loeve transform.

5. The method according to claim 3, wherein the transformation matrix L is:

$\begin{array}{ccc}L=& 1& 1\\ & 1& -1\end{array}.$

6. The method according to claim 1, further comprising the step of:

before the dividing step, reducing a lighting condition asymmetry of the image.

7. The method according to claim 1, further comprising the step of:

before the dividing step, reducing a pose asymmetry of the image.

8. The method according to claim 1, wherein the reduced image data further comprises encoded additional data.

9. The method according to claim 8, wherein the additional data is at least one of text and biometric data.

10. The method according to claim 1, further comprising the step of:

storing the reduced image data on storage media.

11. The method according to claim 10, wherein the storage media is one of a magnetic media, optical media and 2D barcode media.

12. The method according to claim 1, wherein the removing step includes the substeps of averaging and interpolating pixels near the visual artifact.

13. The method according to claim 1, wherein the substantially symmetric subject represents a human face.

14. The method according to claim 1, wherein the image compression algorithm is one of a FFT/Cosine transform algorithm and a wavelet transform algorithm.

15. The method according to claim 1, further comprising the steps of:

computing a line of symmetry of the image; and

dividing the image along the line of symmetry.

16. The method according to claim 1, wherein the weighted average value A and the weighted variance value B are linear functions of the pixel values.

17. A computer-implemented method for reducing an image representing a substantially symmetric subject, comprising the steps of:

dividing the image into a first part and a second part, the second part being substantially symmetric to the first part;

determining pairs of symmetrically corresponding pixels, a member of the pair in the first part having an X pixel value and a member of the pair in the second part having a Y pixel value;

computing a weighted average value A of the pixel values of each one of the pairs;

computing a weighted variance value B of the pixel values of each one of the pairs;

representing each of the pairs with the weighted average and the weighted variance values;

generating reduced image data including at least a plurality of A values and corresponding B values the pairs;

reconstituting the image by computing the X pixel value and the Y pixel value from the weighted average value A and the weighted variance value B; and

compressing the reduced image data with an image compression algorithm, wherein the reconstituting step further includes the substep of removing visual artifacts from the reconstituted image, wherein the removing step includes the substeps of comparing the weighted variance value to a difference in weighted average values for adjacent pixels, wherein the dividing, determining, computing, representing and generating steps are executed by a microprocessor.