BLOCK NOISE DETECTION AND FILTERING

• US 20100202262A1
• Filed: 02/10/2010
• Published: 08/12/2010
• Est. Priority Date: 02/10/2009
• Status: Active Grant

• Current Assignee: Lattice Semiconductor Corporation
• Original Assignee: Anchor Bay Technologies Inc (Lattice Semiconductor Corporation)
• Inventors: Adams, Dale Richard, Thompson, Laurence A.
• Granted Patent: US 8,452,117 B2
• Time in Patent Office: Days
• US Class Current: 369/47.150
• CPC Class Codes: G06T 5/10 using non-spatial domain filtering
• CPC Class Codes: G06T 5/70 Denoising; Smoothing
• CPC Class Codes: G11B 20/00007 Time or data compression or expansion audio compression based on psychoacoustics G10L19/00; data processing for reproducing audio data at different playback speeds G10L21/04; video compression H04N19/00; data compression per se H03M7/30
• CPC Class Codes: H04N 19/117 Filters, e.g. for pre-processing or post-processing sub-band filter banks H04N19/635
• CPC Class Codes: H04N 19/136 Incoming video signal characteristics or properties
• CPC Class Codes: H04N 19/865 with detection of the former encoding block subdivision in decompressed video

Abstract

Systems and methods for block noise detection and filtering are disclosed. One embodiment includes, computing difference magnitudes in pixel values for adjacent pixels in the image. The difference magnitudes can include horizontal difference magnitudes for horizontally adjacent pixels and vertical difference magnitudes for vertically adjacent pixels. One embodiment further includes using normalized sums of the difference magnitudes to determine a set of noise characteristics of the block noise and a set of image characteristics of the image and configuring inputs to the block noise filter using the set of noise and image characteristics.

Claims

1. A method of detecting block noise in an image, the method, comprising:

computing difference magnitudes in pixel values for adjacent pixels in the image;

eliminating a set of difference magnitudes from the difference magnitudes resulting from features in the image;

applying a median filter to the image to generate an output;

subtracting the output from the difference magnitudes;

performing edge detection in the image to detect edge pixel locations.

2. The method of claim 1,

wherein, the difference magnitudes comprise horizontal difference magnitudes for horizontally adjacent pixels and vertical difference magnitudes for vertically adjacent pixels;

further comprising,

averaging the horizontal difference magnitudes in a vertical direction to generate a first vector of average values of the horizontal difference magnitudes of a width of the image;

averaging the vertical difference magnitudes in a horizontal direction to generate a second vector of average values of the vertical difference magnitudes of a height of the image.

3. The method of claim 2, further comprising,

plotting the first vector against pixel locations along the width of the image in a graph.

4. The method of claim 2, further comprising, plotting the second vector against pixel locations along the height of the image in a graph.

5. The method of claim 2, further comprising, removing a DC level from the average values of the vertical difference magnitudes or the average values of the horizontal difference magnitudes.

6. The method of claim 5, wherein, the DC level is removed by applying a median filter to the average values and subtracting the result from the average values.

7. The method of claim 2, further comprising, suppressing values in the first and second vectors near the edges of the image.

8. The method of claim 2, further comprising,

determining normalized sums of the difference magnitudes for a set of candidate interval sizes and a set of candidate offsets;

using a largest value of the normalized sums to determine an interval size and offset of the block noise.

9. The method of claim 8, further comprising, using the interval size and offset to determine a nominal block size and a starting boundary block location.

10. The method of claim 8, further comprising,

determining three normalized sums of the difference magnitudes;

wherein, in a first summation of the three normalized sum, computing sum values using a nominal location and the nominal location −

1 and selecting the location that yields a larger of the sum values;

wherein, in a second summation of the three normalized sum, computing sum values using a nominal location and the nominal location +1 and selecting the location that yields a larger of the sum values;

wherein, in a third summation of the three normalized sum, computing a sum value using the nominal location selecting the nominal location.

11. The method of claim 10, further comprising,

identifying a candidate interval size and offset value which generates a max normalized sum of difference magnitude;

compare the max normalized sum to sums computed for the other two offset values of the candidate interval size to determine a block size of the block noise and to select a final block size and offset value using the nominal block size and the starting boundary location.

12. The method of claim 9,

further comprising, identifying a subsequent boundary block location;

wherein, the subsequent boundary block location is determined by;

incrementing the starting block location by the nominal block size;

identifying a location of a largest or near-largest difference magnitude value in a window around the subsequent boundary block location, and

selecting the location of the largest or near-largest value as the subsequent block boundary location.

13. The method of claim 12, further comprising, generating a block boundary map using multiple subsequent boundary block locations.

14. The method of claim 1, further comprising, using the difference magnitudes to determine whether the image is scaled.

15. The method of claim 1, wherein, the horizontally and vertically adjacent pixels are of 1-pixel differentials.

16. The method of claim 1, further comprising, attenuating the set of difference magnitudes by identifying the set of difference magnitudes equal to or exceeding a threshold value and setting the set of difference magnitudes to the threshold value.

17. The method of claim 1, wherein, the subtracting the output of the median filter from the difference magnitudes further comprises:

clipping values less than zero at zero.

18. The method of claim 1, wherein, the edge detection is performed using Sobel filters.

19. The method of claim 18, wherein, the Sobel filters are 3×

3.

20. The method of claim 2, wherein, the suppressing the edge difference magnitudes at the edge pixel locations further comprises:

setting the edge difference magnitude to zero at a location of the edge pixel locations where a corresponding gradient magnitude equals or exceeds a gradient threshold value.

21. The method of claim 2, wherein, the difference magnitudes are computed for luma and chroma.

22. The method of claim 21, further comprising, generating a chroma block boundary map using the difference magnitudes computed for the chroma and a luma block boundary map using the difference magnitudes computed for the luma.

23. The method of claim 8,

further comprising, using normalized sums to determine a set of noise characteristics of the block noise and using difference magnitude data to determine a set of image characteristics of the image.

24. The method of claim 23,

wherein, the set of noise characteristics include, strength of block noise and block size;

wherein, the set of image characteristics include, amount of detail in the image.

25. The method of claim 23, further comprising, determining a gain value for a block noise filter using the set of noise and image characteristics.

48. The apparatus of claim 5, wherein, the optical disk reader is an HD-DVD reader.

26. A method for configuring a block noise filter for filtering block noise in an image, the method, comprising:

computing difference magnitudes in pixel values for adjacent pixels in the image;

wherein, the difference magnitudes comprise horizontal difference magnitudes for horizontally adjacent pixels and vertical difference magnitudes for vertically adjacent pixels;

using normalized sums of the difference magnitudes to determine a set of noise characteristics of the block noise;

using difference magnitude data to determine a set of image characteristics of the image;

configuring inputs to the block noise filter using the set of noise and image characteristics.

27. The method of claim 26, wherein, the inputs to the block noise filter comprise a gain value having a vertical filter gain and a horizontal filter gain.

28. The method of claim 26,

wherein, the set of noise characteristics include, strength of block noise and block size;

wherein, the set of image characteristics include, amount of detail in the image.

29. The method of claim 26, wherein, the gain value includes a horizontal filter gain and a vertical filter gain.

30. The method of claim 27, further comprising, increasing the gain value of the block noise filter with increasing strength of block noise.

31. The method of claim 27, further comprising, increasing the gain value of the block noise filter with a low amount of detail in the image.

32. The method of claim 27, further comprising, increasing the gain value of the block noise filter with larger block size.

33. The method of claim 26, wherein, the inputs to the block noise filter further comprise an edge gradient multiplier having a horizontal edge multiplier and a vertical edge multiplier.

34. The method of claim 33, further comprising, decreasing the edge gradient multiplier with increasing strength of block noise.

35. The method of claim 33, further comprising, increasing the edge gradient multiplier with a high amount of detail in the image.

36. The method of claim 26,

further comprising, generating a block boundary map of block noise boundary locations in the image;

wherein, the block boundary map includes horizontal and vertical boundaries.

37. The method of claim 26, further comprising,

configuring the block noise filter with the block boundary map, the gain value, and the edge gradient multiplier;

applying the block noise filter to the image to generate a filtered image.

38. The method of claim 37, further comprising, performing edge-blending on the filtered image.

39. The method of claim 37, further comprising, performing edge-smoothing on the filtered image.

40. A method for applying a block noise filter to an image, the method, comprising:

selecting a gain setting for the block noise filter based on block noise characteristics in the image;

wherein, the gain setting includes a vertical gain factor;

computing a set of coefficients for a vertical luma filter using the vertical gain factor;

computing a set of coefficients for a vertical chroma filter using the vertical gain factor;

filtering the image along a vertical dimension using the coefficients for the vertical and chroma filters to generate a vertically-filtered image.

41. The method of claim 40,

wherein, the gain setting includes a horizontal gain factor;

further comprising, computing a set of coefficients for a horizontal luma filter using the horizontal gain factor;

computing a set of coefficients for a horizontal chroma filter using the horizontal gain factor;

filtering the vertically-filtered image along a horizontal dimension using the coefficients for the horizontal and chroma filters to generate a horizontally-filtered image.

42. The method of claim 41, further comprising,

generating a first blended image by blending the vertically-filtered image and the image;

generating a second blended image by blending the vertically-filtered image and the horizontally filtered image.

43. The method of claim 42, further comprising, performing edge-smoothing on the blended image.

44. A system, comprising:

means for, computing difference magnitudes in pixel values for adjacent pixels in the image;

means for, eliminating a set of difference magnitudes from the difference magnitudes resulting from features in the image;

means for, determining normalized sums of the difference magnitudes for a set of candidate interval sizes and a set of candidate offsets;

means for, computing determining averages of rows and columns of the difference magnitude values;

means for, using the interval size and offset to determine a nominal block size and a starting boundary block location;

means for, identifying a subsequent boundary block location;

wherein, the subsequent boundary block location is determined by incrementing the starting block location using the nominal block size;

means for, using a largest value of the normalized sums to determine an interval size and offset of the block noise;

means for, generating a block boundary map using multiple subsequent boundary block locations.

45. An optical disk apparatus, comprising:

an optical disk reader, wherein, when in operation, reads an optical disk;

a memory coupled to the optical disk reader; and

a graphics processor coupled to the memory, wherein the graphics processor executes and algorithm which applies a block noise filter to an image by;

selecting a gain setting for the block noise filter based on block noise characteristics in the image;

wherein, the gain setting includes a vertical gain factor and a horizontal gain factor;

computing coefficients for a vertical luma filter using a vertical gain factor;

computing coefficients for a vertical chroma filter using the vertical gain factor;

filtering the image along a vertical dimension using the coefficients for the vertical and chroma filters to generate a vertically-filtered image;

computing coefficients for a horizontal luma filter using the horizontal gain factor;

computing coefficients for a horizontal chroma filter using the horizontal gain factor;

filtering the vertically-filtered image along a horizontal dimension using the coefficients for the horizontal and chroma filters to generate a horizontally-filtered image;

generating a first blended image by blending the vertically-filtered image and the image;

generating a second blended image by blending the vertically-filtered image and the horizontally-filtered image.

46. The apparatus of claim 45, wherein, the optical disk reader is a Blu-ray disk reader.

47. The apparatus of claim 45, wherein, the optical disk reader is a DVD reader.