Methods to estimate noise variance from a video sequence

• US 7,715,645 B2
• Filed: 11/17/2004
• Issued: 05/11/2010
• Est. Priority Date: 11/17/2004
• Status: Expired due to Fees

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Zhou, Zhi, Kim, Yeong-Taeg
• Primary Examiner(s): Kassa; Yosef
• Application Number:
  US10/991,265
• Publication Number:
  US 20060103765A1
• Time in Patent Office: 2,001 Days
• Field of Search: 382/209, 382/219, 382/252, 382/274, 382/275, 358/3.26, 358/3.27, 358/463, 348/14.15, 348/208.4
• US Class Current: 382/254
• CPC Class Codes: H04N 5/21 Circuitry for suppressing or minimising disturbance, e.g. moiré or halo

Abstract

A noise estimation system that estimates noises in a sequence of video frames, uses a an image structure remover and a noise variance calculator. The image structure remover computes the difference over a local window in two consecutive frames. The noise variance calculator estimates the noise variance (standard deviation) from the distribution of the local difference. If there is no motion or small motion between two consecutive frames, the image structure can be removed by calculating the local difference, whereby very robust estimation is obtained.

Claims

1. A method of video noise estimation for a sequence of video image frames, comprising:

determining local differences between a portion of a frame and a corresponding portion of a subsequent frame;

determining a statistical distribution of the local differences; and

,

based on the statistical distribution, estimating a noise variance.

2. The method of claim 1, wherein determining the difference comprises computing a mean absolute error for the two frames, wherein determining the statistical distribution comprises determining a histogram of the local differences.

3. The method of claim 2, wherein estimating the noise variance further comprises determining the value associated with the maximum frequency of the histogram.

4. The method of claim 3, wherein determining the difference comprises computing a mean square error for the two frames.

5. The method of claim 3, wherein determining the difference comprises computing a sum of square error for the two frames.

6. The method of claim 3, wherein determining the difference comprises computing a difference of local mean for the two frames.

7. The method of claim 3, wherein determining the difference comprises determining the difference of two matching blocks in the two frames, respectively, wherein the two matching blocks can be searched by motion estimation.

8. The method of claim 7, wherein estimating the noise variance comprises estimating a first order expectation for the distribution.

9. The method of claim 7, wherein estimating the noise variance comprises estimating a variance for the distribution.

10. The method of claim 7, wherein estimating the noise variance comprises estimating a maximum histogram position for the distribution.

11. The method of claim 10, wherein the two frames are consecutive frames.

23. The method of claim 11, wherein the frame and the subsequent frame are actual frames.

24. The method of claim 23, wherein determining a difference between a portion of a frame and a corresponding portion of a subsequent frame comprises determining a local difference between a pixel in the frame and a corresponding pixel in the subsequent frame.

25. The method of claim 24, wherein estimating a noise variance as a function of a statistical distribution of the difference further comprises compensating for the effects of motion between the frame and the subsequent frame.

12. A video noise estimation system that estimates video noise for a sequence of video image frames, comprising:

an image structure remover that determines local differences between a portion of a frame and a corresponding portion of a subsequent frame; and

a noise estimator that determines a statistical distribution of the local differences and estimates the noise variance based on the statistical distribution.

13. The system of claim 12, wherein the statistical distribution further comprises a histogram of the local differences, wherein the image structure remover further determines the difference by computing a mean absolute error for the two frames.

14. The system of claim 13, wherein estimating the noise variance further comprises determining the value associated with the maximum frequency of the histogram, wherein the image structure remover further determines the difference by computing a sum of absolute error for the two frames.

15. The system of claim 13, wherein the image structure remover further determines the difference by computing a mean square error for the two frames.

16. The system of claim 13, wherein the image structure remover further determines the difference by computing a sum of square error for the two frames.

17. The system of claim 13, wherein the image structure remover further determines the difference by computing a difference of local mean for the two frames.

18. The system of claim 17, wherein the image structure remover determines the difference of two matching blocks in the two frames, respectively, wherein the two matching blocks can be searched by motion estimation.

19. The system of claim 17, wherein the noise estimator further estimates the noise variance by estimating a first order expectation for the distribution.

20. The system of claim 17, wherein the noise estimator further estimates the noise variance by estimating a variance for the distribution.

21. The system of claim 17, wherein the noise estimator further estimates the noise variance by estimating a maximum histogram position for the distribution.

22. The system of claim 21, wherein the two frames are consecutive frames.

26. A method of video noise estimation for a sequence of video image frames, comprising:

determining a local noise difference between a local window of a frame and a corresponding local window of a successive frame;

estimating a noise variance as a function of a statistical histogram of the local noise difference,

determining a motion value between the two successive frames based on the noise variance; and

removing an image structure if the motion value is less than a predetermined motion value.

27. The method of claim 26, wherein removing the image structure is based on calculating the local noise difference.

28. A video noise estimation system that estimates video noise for a sequence of video image frames, comprising:

an image structure remover that determines a local noise difference between a window of a frame and a corresponding window of a successive frame;

a noise estimator that estimates the noise variance as a function of the statistical histogram of the local noise difference; and

a motion estimator configured to estimate a motion value between the two successive frames,

wherein the image structure remover removes an image structure upon the motion value being less than a predetermined motion value.

29. The system of claim 28, wherein the image structure remover removes the image structure based on calculating the local noise difference.