Reversible diffusion-based compression

• US 7,706,448 B2
• Filed: 07/30/2003
• Issued: 04/27/2010
• Est. Priority Date: 07/30/2003
• Status: Expired due to Fees

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Varma, Ankur
• Primary Examiner(s): VO, TUNG T
• Application Number:
  US10/630,955
• Publication Number:
  US 20050025245A1
• Time in Patent Office: 2,463 Days
• Field of Search: 375/240.23, 375/240.12, 375/240.13, 375/240.14, 375/240.25
• US Class Current: 375/240.23
• CPC Class Codes: H04N 19/85 using pre-processing or post-processing specially adapted for video compression
• CPC Class Codes: H04N 19/91 Entropy coding, e.g. variable length coding [VLC] or arithmetic coding

Abstract

Subject matter includes exemplary methods of reversible diffusion-based compression and an exemplary compression engine. In one implementation, a reversible diffusion function is applied to decrease high spatial frequency pixel values in an image or a prediction error image residue and to smooth variances between adjacent pixel values. An exemplary reversible diffusion function can increase data compression without loss of high frequency information yet operate with online encoders and decoders that lack significant processing power. An exemplary method transforms the data to make the data more amenable to compression schemes that utilize entropy transforms as an intermediate processing step, for example, prior to Huffman coding.

Claims

1. A noise reduction engine, comprising:

a buffer for storing pixel values;

a matrix selector for selecting dimensions of matrices for arranging the pixel values to represent regions of prediction error image residue;

a diffusion engine for applying a reversible diffusion function for reducing the magnitude of at least some of the pixel values and for reducing variability in the difference between adjacent pixel values in a subject matrix by diffusing magnitudes of pixel values into each other to thereby reduce noise associated with application of a frequency domain transform and application of an inverse frequency domain transform, wherein an anchor value facilitates reverse diffusion by the reversible diffusion function; and

an anchor value selector associated with the diffusion engine to select one of the pixel values in a given matrix as the anchor value, wherein the anchor value facilitates the reverse diffusion by the reversible diffusion function to restore the magnitude of the at least some of the pixel values after application of the frequency domain transform to the diffused pixel values and application of the inverse frequency domain transform to recover the diffused pixel values.

2. The noise reduction engine as recited in claim 1, further comprising an entropy calculator associated with the anchor value selector to select an anchor value based on an entropy value of one or more of the pixel values.

3. The noise reduction engine as recited in claim 1, further comprising a scan pattern engine to apply the reversible diffusion function to a matrix of pixel values in an order.

4. The noise reduction engine as recited in claim 3, further comprising an iteration manager to control an amount of diffusion to be applied to a matrix of pixel values by controlling a number of times that the reversible diffusion function is applied.

5. The noise reduction engine as recited in claim 1, further comprising a store of reversible diffusion functions suitable for different image residues.

6. The noise reduction engine as recited in claim 1, further comprising a reverse diffusion module to apply the reverse diffusion using the anchor value.

7. One or more computer storage media having computer-executable instructions stored thereon that, when executed, direct at least one computer to perform acts, the acts comprising:

storing pixel values in a buffer;

selecting, by a matrix selector, dimensions of matrices for arranging the pixel values to represent regions of prediction error image residue

applying, by a diffusion engine, a reversible diffusion function for reducing the magnitude of at least some of the pixel values and for reducing variability in the difference between adjacent pixel values in a subject matrix by diffusing magnitudes of pixel values into each other to thereby reduce noise associated with application of a frequency domain transform and application of an inverse frequency domain transform, wherein an anchor value facilitates reverse diffusion by the reversible diffusion function; and

selecting, by an anchor value selector associated with the diffusion engine, one of the pixel values in a given matrix as the anchor value, wherein the anchor value facilitates the reverse diffusion by the reversible diffusion function to restore the magnitude of the at least some of the pixel values after application of the frequency domain transform to the diffused pixel values and application of the inverse frequency domain transform to recover the diffused pixel values.

8. The one or more computer storage media as recited in claim 7, the acts further comprising selecting, by an entropy calculator associated with the anchor value selector, an anchor value based on an entropy value of one or more of the pixel values.

9. The one or more computer storage media as recited in claim 7, the acts further comprising applying, by a scan pattern engine, the reversible diffusion function to a matrix of pixel values in an order.

10. The one or more computer storage media as recited in claim 9, the acts further comprising controlling, by an iteration manager, an amount of diffusion to be applied to a matrix of pixel values by controlling a number of times that the reversible diffusion function is applied.

11. The one or more computer storage media as recited in claim 7, the acts further comprising maintaining a store of reversible diffusion functions suitable for different image residues.

12. The one or more computer storage media as recited in claim 7, the acts further comprising applying, by a reverse diffusion module, the reverse diffusion using the anchor value.

13. A system, comprising:

at least one memory;

at least one processor;

a buffer for storing pixel values;

instructions stored in the at least one memory and executed by the at least one processor, the instructions comprising;

a matrix selector for selecting dimensions of matrices for arranging the pixel values to represent regions of prediction error image residue;

a diffusion engine for applying a reversible diffusion function for reducing the magnitude of at least some of the pixel values and for reducing variability in the difference between adjacent pixel values in a subject matrix by diffusing magnitudes of pixel values into each other to thereby reduce noise associated with application of a frequency domain transform and application of an inverse frequency domain transform, wherein an anchor value facilitates reverse diffusion by the reversible diffusion function; and

an anchor value selector associated with the diffusion engine to select one of the pixel values in a given matrix as the anchor value wherein the anchor value facilitates the reverse diffusion by the reversible diffusion function to restore the magnitude of the at least some of the pixel values after application of the frequency domain transform to the diffused pixel values and application of the inverse frequency domain transform to recover the diffused pixel values.

14. The system as recited in claim 13, the instructions further comprising an entropy calculator associated with the anchor value selector to select an anchor value based on an entropy value of one or more of the pixel values.

15. The system as recited in claim 13, the instructions further comprising a scan pattern engine to apply the reversible diffusion function to a matrix of pixel values in an order.

16. The system as recited in claim 15, the instructions further comprising an iteration manager to control an amount of diffusion to be applied to a matrix of pixel values by controlling a number of times that the reversible diffusion function is applied.

17. The system as recited in claim 13, the instructions further comprising a store of reversible diffusion functions suitable for different image residues.

18. The noise reduction engine as recited in claim 13, the instructions further comprising a reverse diffusion module to apply the reverse diffusion using the anchor value.