Method of removing block boundary noise components in block-coded images

• US 6,741,752 B1
• Filed: 04/17/2000
• Issued: 05/25/2004
• Est. Priority Date: 04/16/1999
• Status: Expired due to Term

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Yang, Jeong-Hun
• Primary Examiner(s): Johnson, Timothy M.
• Assistant Examiner(s): Choobin, Barry
• Application Number:
  US09/550,728
• Time in Patent Office: 1,499 Days
• Field of Search: 382/268, 382/250, 382/199, 382/253, 382/254, 382/200, 263/266, 263/261, 263/197, 263/239, 358/426-1-4, 358/1.19, 375/240.03, 375/240.24, 348/606
• US Class Current: 382/268
• CPC Class Codes: G06T 2207/20021 Dividing image into blocks, subimages or windows
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: G06T 5/70 Denoising; Smoothing
• CPC Class Codes: G06T 9/005 Statistical coding, e.g. Huffman, run length coding
• CPC Class Codes: H04N 19/86 involving reduction of coding artifacts, e.g. of blockiness

Abstract

An improved block noise elimination method is provided, which uses one-dimensional pixel vectors made of pixel rows or columns across two neighboring blocks to analyze blocky noise residing across the neighboring blocks. The blocky noise in each pixel-vector is modeled as shape vector multiplied by boundary discontinuity. The shape vector, varying with local image activity, is estimated by Minimum Mean Squared Error (MMSE) approach. Depending upon the shape vector, the corresponding blocky noise is eliminated in postprocessing of decoded images so that encoded images may be restored to their original images.

Claims

1. A method for restoring a block-encoded image, comprising:

estimating block boundary noise components residing across two neighboring blocks in the block-encoded image, wherein the block boundary noise components comprise block boundary discontinuity parameters, which are estimated by subtracting the sum of the average of internal pixel slopes divided by two from a boundary pixel slope;

decoding the block-encoded image to obtain a decoded image; and

eliminating the estimated block boundary noise components from the decoded image.

2. The method of claim 1, wherein said block boundary noise components are statistically estimated from encoded pixel vectors of plural images encoded at several bit rates.

3. The method of claim 2, wherein said block boundary noise components are estimated using said block boundary discontinuity parameters and local image activity parameters of the encoded pixel vectors.

4. The method of claim 3, wherein said block boundary discontinuity parameters are slopes of the pixel values across block boundary.

5. The method of claim 3, wherein said local image activity parameters are nonzero coefficient indices of highest frequency in N-point DCT (Discrete Cosine Transform) coefficient vectors of the respective pixel vectors.

6. The method of claim 3, wherein said block boundary noise components are denoted by shape vectors representing shapes of block boundary noises across adjacent blocks.

7. The method of claim 6, wherein said shape vectors are one-dimensional vectors corresponding to pixel rows or columns of the adjacent blocks.

8. The method of claim 6, wherein said shape vectors are estimated with respect to the local image activity parameters.

9. The method of claim 8, wherein said shape vectors are estimated by minimum mean squared error (MMSE) from statistics of the encoded pixel vectors.

10. The method of claim 7, wherein said shape vectors are estimated using original pixel vectors of the block-encoded image, the encoded pixel vectors and the block boundary discontinuity parameters of the encoded pixel vectors.

11. The method of claim 3, wherein said eliminating the estimated block boundary noise components from the decoded image comprises:

calculating the block boundary discontinuity parameters and the local image activity parameters of the encoded pixel vectors;

obtaining shape vectors corresponding to the respective local image activity parameters; and

removing products of the block boundary discontinuity parameters and the shape vectors from decoded pixel vectors, respectively.

12. The method of claim 1, further comprising transferring the estimated block boundary noise components to a decoding part prior to said eliminating the estimated block boundary noise components.

13. The method of claim 12, wherein said estimated block boundary noise components are transferred to the decoding part together with encoded pixel vectors in real time.

14. The method of claim 12, wherein said estimated block boundary noise components are previously stored in a table within the decoding part.

15. A method for restoring encoded images, comprising:

sequentially receiving encoded pixel vectors of the encoded images;

calculating local image activities of the encoded pixel vectors, each local image activity corresponding to each encoded pixel vector;

estimating shape vectors representing block boundary noises of the encoded images, the shape vectors varying with the local image activities and each shape vector corresponding to each encoded pixel vector;

sequentially receiving decoded pixel vectors of decoded images;

calculating local image activities of the decoded pixel vectors, each local image activity corresponding to each decoded pixel vector;

obtaining the shape vectors estimated in the step of estimating, each obtained shape vector corresponding to each local image activity; and

removing each of the shape vectors obtained in the step of obtaining from corresponding one of the decoded pixel vectors.

16. The method of claim 15, wherein the step of estimating shape vectors includes the steps of:

calculating block boundary discontinuities of the respective encoded pixel vectors;

obtaining boundary noise components of the respective encoded pixel vectors, each boundary noise component being obtained by dividing a difference between each encoded pixel vector and a corresponding original pixel vector by the corresponding block boundary discontinuity; and

adding each of the boundary noise components to a previously estimated shape vector.

17. The method of claim 15, wherein the step of removing the shape vectors includes the steps of:

calculating block boundary discontinuities of the respective decoded pixel vectors;

obtaining products of the respective block boundary discontinuities of the respective decoded pixel vectors and the respective shape vectors obtained in the step of obtaining; and

subtracting each of the products from corresponding one of the decoded pixel vectors.

18. The method of claim 15, wherein one of the local image activities corresponding to one of the pixel vectors is determined by a largest one of nonzero coefficient indices of the corresponding pixel vector.

19. The method of claim 15, wherein the step of obtaining the shape vectors includes retrieving corresponding shape vectors which are previously stored in a table of a decoding part.