DYNAMIC NOISE FILTER AND SIGMA FILTERING METHOD

• US 20100188582A1
• Filed: 03/05/2009
• Published: 07/29/2010
• Est. Priority Date: 01/23/2009
• Status: Active Grant

• Current Assignee: Silicon Integrated Systems Corporation (Silicon Integrated Systems Corporation Taiwan)
• Original Assignee: Silicon Integrated Systems Corporation (Silicon Integrated Systems Corporation Taiwan)
• Inventors: HSU, Jeng-Yun
• Granted Patent: US 8,175,414 B2
• Time in Patent Office: Days
• US Class Current: 348/607
• CPC Class Codes: H04N 5/145 Movement estimation for video coding H04N19/51
• CPC Class Codes: H04N 5/213 Circuitry for suppressing or minimising impulsive noise for suppressing or minimising disturbance in image signal generation H04N23/81

Abstract

A dynamic noise filter is disclosed, which includes a noise estimation circuit, a TNR filtering circuit, a SNR filtering circuit, a blending circuit and a motion estimation circuit. By performing noise estimation for each image, the invention dynamically adjusts the strengths of the TNR filter and the SNR filter as well as a blending ratio between the outputs of the TNR filter and the SNR filter, thereby obtaining the best image quality.

Claims

1. A dynamic noise filter, comprising:

a noise estimation circuit for receiving a current image and its corresponding preceding image and performing noise estimation to generate a temporal attribute parameter and a spatial attribute parameter;

a temporal noise reduction (TNR) filtering circuit for receiving the current image and its corresponding preceding image and modifying its own TNR characteristic curve to generate at least one first filtered scan line and the corresponding first parameter flow according to the temporal attribute parameter generated by the noise estimation circuit at the preceding image period;

a motion estimation circuit for performing both a setting and a motion estimation of a mosquito noise block according to an input image and its corresponding preceding image;

a spatial noise reduction (SNR) filtering circuit for generating a second filtered scan line and a corresponding second parameter flow according to the spatial attribute parameter generated by the noise estimation circuit at the preceding image period, preceding outputs of the motion estimation circuit, and the at least one first filtered scan line and the corresponding first parameter flow; and

a blending circuit for sequentially blending the second filtered scan line and one of the at least one first filtered scan line to generate an output scan line according to the corresponding second parameter flow.

2. The filter according to claim 1, further comprising:

a first delay circuit, coupled among the noise estimation circuit, the TNR filtering circuit, the SNR filtering circuit, for delaying the temporal attribute parameter and the spatial attribute parameter for one image period; and

a second delay circuit, coupled between the motion estimation circuit and the SNR filtering circuit, for delaying outputs of the motion estimation circuit for one or two field periods when the current image is a field and for delaying the outputs of the motion estimation circuit for one frame period when the current image is a frame.

3. The filter according to claim 1, wherein the input image is the current image or the output image of the TNR filtering circuit.

4. The filter according to claim 3, wherein, when the input image is the output of the TNR filtering circuit, the front end of the TNR filtering circuit is further provided with a third delay circuit for delaying the corresponding preceding image of the input image for the operating time of the TNR filtering circuit.

5. The filter according to claim 1, wherein the spatial attribute parameter comprises hi coring, hc_coring, vl_coring, and vc_coring.

6. The filter according to claim 5, wherein the SNR filtering circuit is implemented using a vertical horizontal filter and the vertical horizontal filter comprises at least one vertical SNR filtering unit and at least one horizontal SNR filtering unit.

7. The filter according to claim 6, wherein each horizontal SNR filtering unit and each vertical SNR filtering unit modify the spatial attribute parameter and the second parameter flow according to the preceding outputs of the motion estimation circuit.

8. The filter according to claim 7, wherein each horizontal filtering unit dynamically modifies its own SNR characteristic curve according to the parameters hl_coring and hc_coring and each vertical filtering unit modifies its own SNR characteristic curve according to the parameters vl_coring and vc_coring.

9. The filter according to claim 1, wherein the SNR filtering circuit comprises:

a vertical horizontal filter for receiving the at least one first filtered scan line and the corresponding first parameter flow and for modifying its own SNR characteristic curve according to the spatial attribute parameter generated by the noise estimation circuit at the preceding image period and the preceding outputs of the motion estimation circuit to thereby generate a third filtered scan line and the second parameter flow, wherein the vertical horizontal filter comprises at least one vertical filtering unit and at least one horizontal filtering unit;

a sigma filter for receiving the at least one first filtered scan line and the corresponding first parameter flow and for performing an average calculation for a block including a target pixel according to a predetermined pixel difference range and a predetermined negative slope to thereby generate a fourth filtered scan line and its corresponding salt-and-pepper flag flow;

a decision circuit for deciding to enable either the vertical horizontal filter or the sigma filter according to a preset edge value and a difference between the maximum luminance value and the minimum luminance value of the block including the target pixel in the at least one first filtered scan line; and

a median filter for receiving the outputs of the vertical horizontal filter and the sigma filter and for deciding whether to perform a median filtering operation over the block including the target pixel to thereby generate the second filtered scan line and its corresponding second parameter flow according to a salt-and-pepper flag and a mosquito flag of the target pixel.

10. The filter according to claim 1, wherein the temporal attribute parameter comprises nr_alpha and nr_th1.

11. The filter according to claim 1, wherein the noise estimation circuit generates the temporal attribute parameter and the spatial attribute parameter according to the noise mean and the noise variance of a plurality of specific pixels in the current image.

12. The filter according to claim 11, wherein the specific pixels are pixels having the several lowest noise levels in the current image.

13. The filter according to claim 11, wherein the noise estimation circuit further modifies the magnitude of the temporal attribute parameter according to the degree that the current image is close to a static image.

14. A dynamic noise filter, comprising:

a noise estimation circuit for receiving a current image and its corresponding preceding image and performing noise estimation to generate a temporal attribute parameter; and

a temporal noise reduction (TNR) filtering coupled to the noise estimation circuit for receiving the current image and its corresponding preceding image and for modifying its own TNR characteristic curve to generate a temporal filtered image according to the temporal attribute parameter generated by the noise estimation circuit at the preceding image period.

15. The filter according to claim 14, further comprising:

a delay circuit coupled to the noise estimation circuit for delaying the temporal attribute parameter for one image period.

16. The filter according to claim 14, wherein the noise estimation circuit generates the temporal attribute parameter according to the noise mean and the noise variance of a plurality of specific pixels in the current image.

17. The filter according to claim 16, wherein the specific pixels are pixels having the several lowest noise levels in the current image.

18. The filter according to claim 16, wherein the noise estimation circuit further modifies the magnitude of the temporal attribute parameter according to the degree that the current image is close to a static image.

19. The filter according to claim 14, wherein the temporal attribute parameter comprises nr_alpha and nr_th1.

20. A dynamic noise filter, comprising:

a noise estimation circuit for receiving a current image and its corresponding preceding image, performing noise estimation and generating a spatial attribute parameter; and

a motion estimation circuit for performing both a setting and a motion estimation of a mosquito noise block over the current image and its corresponding preceding image; and

a spatial noise reduction (SNR) filtering circuit for receiving the current image and generating a spatial filtered image according to the spatial attribute parameter generated by the noise estimation circuit at the preceding image period and preceding outputs of the motion estimation circuit.

21. The filter according to claim 20, wherein the noise estimation circuit generates the spatial attribute parameter according to the noise mean and the noise variance of a plurality of specific pixels in the current image.

22. The filter according to claim 21, wherein the specific pixels are pixels having the several lowest noise levels in the current image.

23. The filter according to claim 20, further comprising:

a first delay circuit coupled between the noise estimation circuit and the SNR filtering circuit for delaying the spatial attribute parameter for one image period; and

a second delay circuit coupled between the motion estimation circuit and the SNR filtering circuit for delaying the outputs of the motion estimation circuit for one or two field periods when the current image is a field and delaying the outputs of the motion estimation circuit for one frame period when the current image is a frame.

24. The filter according to claim 20, wherein the spatial attribute parameter comprises hl_coring, hc_coring, vl_coring, and vc_coring.

25. The filter according to claim 24, wherein the SNR filtering circuit is implemented using a vertical horizontal filter and the vertical horizontal filter comprises at least one vertical SNR filtering unit and at least one horizontal SNR filtering unit.

26. The filter according to claim 25, wherein each horizontal filtering unit and each vertical filtering unit modify the spatial attribute parameter according to the preceding outputs of the motion estimation circuit.

27. The filter according to claim 26, wherein each horizontal filtering unit dynamically modifies its own SNR characteristic curve according to the parameters hl_coring and hc_coring and each vertical filtering unit dynamically modifies its own SNR characteristic curve according to the parameters vl_coring and vc_coring.

28. A sigma filtering method, for processing at least one image, the method comprising the following steps:

receiving a m×

n block of the image;

according to a target pixel of the m×

n block, calculating a actual pixel difference for each of the other pixels in the m×

n block;

modifying each actual pixel difference to be a corresponding modified pixel difference according to a predetermined pixel difference range and a predetermined negative slope; and

obtaining a filtered output value according to each modified pixel difference, the target pixel value and the number of pixels in the m×

n block.

29. The method according to claim 28, wherein the predetermined pixel difference range is from (−

R) to R and R=ppdiff×

factor, where ppdiff is a difference between the maximum pixel value and the minimum pixel value in the m×

n block and factor is a predetermined value.

30. The method according to claim 29, wherein the step of modifying further comprises:

calculating the modified pixel difference according to the R value and the predetermined negative slope while the actual pixel difference is more than R; and

setting the modified pixel difference to 0 if the modified pixel difference is less than 0.

31. The method according to claim 29, wherein the step of modifying further comprises:

calculating the modified pixel difference according to the (−

R) value and the predetermined negative slope while the actual pixel difference is less than (−

R); and

setting the modified pixel difference to 0 if the modified pixel difference is more than 0.

32. The method according to claim 29, wherein the pixel difference is a luminance difference or a chrominance difference.