High quality wide-range multi-layer image compression coding system

• US 20060071825A1
• Filed: 09/12/2005
• Published: 04/06/2006
• Est. Priority Date: 09/14/2004
• Status: Active Grant

• Current Assignee: Gary Demos
• Original Assignee: Gary Demos
• Inventors: Demos, Gary
• Granted Patent: US 7,916,952 B2
• Time in Patent Office: Days
• US Class Current: 341/50
• CPC Class Codes: H03M 7/24 Conversion to or from floating-point codes
• CPC Class Codes: H04N 19/10 using adaptive coding
• CPC Class Codes: H04N 19/115 Selection of the code volume for a coding unit prior to coding
• CPC Class Codes: H04N 19/124 Quantisation
• CPC Class Codes: H04N 19/146 Data rate or code amount at the encoder output
• CPC Class Codes: H04N 19/162 User input
• CPC Class Codes: H04N 19/17 the unit being an image region, e.g. an object
• CPC Class Codes: H04N 19/187 the unit being a scalable video layer
• CPC Class Codes: H04N 19/29 involving scalability at the object level, e.g. video object layer [VOL]
• CPC Class Codes: H04N 19/61 in combination with predictive coding
• CPC Class Codes: H04N 19/63 using sub-band based transform, e.g. wavelets
• CPC Class Codes: H04N 19/635 characterised by filter definition or implementation details
• CPC Class Codes: H04N 19/85 using pre-processing or post-processing specially adapted for video compression

Abstract

Systems, methods, and computer programs for high quality wide-range multi-layer image compression coding, including consistent ubiquitous use of floating point values in essentially all computations; an adjustable floating-point deadband; use of an optimal band-split filter; use of entire SNR layers at lower resolution levels; targeting of specific SNR layers to specific quality improvements; concentration of coding bits in regions of interest in targeted band-split and SNR layers; use of statically-assigned targets for high-pass and/or for SNR layers; improved SNR by using a lower quantization value for regions of an image showing a higher compression coding error; application of non-linear functions of color when computing difference values when creating an SNR layer; use of finer overall quantization at lower resolution levels with regional quantization scaling; removal of source image noise before motion-compensated compression or film steadying; use of one or more full-range low bands; use of alternate quantization control images for SNR bands and other high resolution enhancing bands; application of lossless variable-length coding using adaptive regions; use of a folder and file structure for layers of bits; and a method of inserting new intra frames by counting the number of bits needed for a motion compensated frame.

Claims

1. A method for compressing a sequence of digitized video images represented as pixels, the method including:

applying an optimal band-split filter phase construction to a digitized video image in which low-pass bands are created in between pixels of the digitized video image.

2. The method of claim 1, wherein at least the lowest low-pass band has a floating-point numeric representation.

3. A method for compressing a sequence of digitized video images, the method including:

applying a band-split filter structure to generate a band-split hierarchy from a digitized video image, the band-split hierarchy comprising at least two reduced resolution levels in which low-pass bands are created in between pixels of the digitized video image;

applying a different quantization parameter value to each of such at least two reduced resolution layers; and

performing subsequent encoding steps on such at least two reduced resolution layers to create a compressed output of such digitized video image.

4. The method of claim 3, further including the step of constraining at least one quantization parameter to a selected extreme value.

5. The method of claim 3, wherein at least one quantization parameter value has a floating-point numeric representation.

6. A method for compressing a sequence of digitized video images, the method including:

transforming each digitized video image to a floating-point numeric representation; and

performing all encoding steps prior to quantization using such floating-point numeric representation to create a compressed output.

7. A method for removing noise from digitized video images, the method including:

statistically characterizing at least one of temporal, spatial, and pattern noise within a source of such digitized video images as a function of at least one of color, brightness, and region; and

removing such characterized noise prior to compression or application of image steadying algorithms.

8. A method of adding signal-to-noise improvement layers which are targeted to specific aspects of a sequence of images, the method including:

selecting one or more aspects of a coded image to improve;

isolating those image aspects using the selection criteria;

coding a signal-to-noise layer using the isolated coded image aspects.

9. The method of claim 8, further including the step of applying one of a high-pass or band-pass filter to select specific spatial frequencies as image aspects.

10. The method of claim 8, wherein the step of selecting further includes selecting a plurality of static regions of the coded image to improve, and coding at least one region differently from another region.

11. A method for determining when to insert an intra frame within a sequence of motion-compensated compressed digitized video images, the method including:

counting a first number of bits needed to encode an intra frame;

counting a second number of bits needed to encode a motion compensated frame subsequent to the intra frame;

comparing the second number to the first number to determine whether to recompute the motion compensated frame as an intra frame.

12. A method for compressing a sequence of digitized video images, the method including:

generating a hierarchy of processed images from a digitized video image;

determining a plurality of regions within each processed image, each region being selected based on lossless coding efficiency; and

applying a lossless variable-length coding independently to each such region.

13. The method of claim 12, wherein the hierarchy of processed images comprises a quantized band-split hierarchy of images generated from the digitized video image.

14. The method of claim 12, wherein the hierarchy of processed images comprises at least one low resolution floating-point low band generated from the digitized video image.

15. The method of claim 12, wherein the hierarchy of processed images includes at least one floating-point regional quantization image generated from the digitized video image.

16. The method of claim 12, wherein the lossless variable-length coding applied to each region is selected based on coding efficiency.

17. A method for grouping data resulting from compressing a sequence of digitized video images, the method including:

storing data related to an organizational aspect of the result of compression of a single image in a file data structure; and

grouping all file data structures related to each aspect within a folder data structure.

18. The method of claim 17, where at least one folder data structure is optionally omitted from decoding of the compressed digitized video images.

19. The method of claim 17, wherein the organizational aspect of the result of compression of a single image includes at least one of a resolution layer, a signal-to-noise layer, a floating-point lowest band, a quantization regional low band, or motion compensation data.

20. A method for decompressing a sequence of compressed digitized video images, the method including:

performing all decoding steps after dequantization using a floating-point numeric representation to create an output comprising decompressed digitized pixel values.

21. The method of claim 20, further including the step of converting the final floating-point numeric representation of such pixel values to integers.

22. A method for compressing a sequence of digitized video images represented as pixels, the method including:

generating at least one processed layer from at least one of the digitized video images, each processed layer comprising a plurality of values; and

applying a quantization parameter to the plurality of values of such at least one processed layer using a lower quantization value for regions of such layer showing a higher compression coding error relative to a corresponding digitized video image or layer from which such processed layer was generated.

23. A method for compressing a sequence of digitized video images represented as pixels, the method including:

generating at least one processed layer from at least one of the digitized video images, each processed layer comprising a plurality of values; and

applying a quantization parameter to the plurality of values of such at least one processed layer using an adjustable size floating-point deadband for values of such processed layer near zero.

24. The method of claim 23, further including the step of adjusting the size of the deadband as a function of at least one of:

applying the quantization parameter to independent regions within one of such processed layers, or applying the quantization parameter to different processed layers.

25. The method of claim 23, wherein at least one processed layer is generated by the bi-orthogonal subband 9/7 discrete wavelet transform.

26. A method for compressing a sequence of digitized video images, the method including:

generating a band-split hierarchy from a digitized video image, the band-split hierarchy comprising at least two reduced resolution levels;

applying a quantization parameter, represented as a one of a floating-point or half floating-point numeric value, to each of such at least two reduced resolution layers; and

performing subsequent encoding steps on such at least two reduced resolution layers to create a compressed output of such digitized video image.

27. The method of claim 26, wherein the band split hierarchy is generated by the bi-orthogonal subband 9/7 discrete wavelet transform.

38. The method of claim 26, wherein the non-linear function is a log function.

28. A method for compressing a sequence of digitized video images represented as pixels, the method including:

generating at least one processed layer from at least one of the digitized video images, each processed layer comprising a plurality of values;

dithering the values of at least one of such processed layers with pseudo-random noise prior to quantization; and

applying a quantization parameter to the plurality of values of such at least one processed layer.

29. The method of claim 28, wherein each value of each processed layer has an inherent noise level, and the quantization parameter has a step size, further including the step of setting the level of the pseudo-random noise such that the sum of the inherent noise and pseudo-random noise for at least each value outside of a deadband around zero approximates the quantization parameter step size.

30. A method for compressing a sequence of digitized video images represented as pixels, the method including:

filtering a digitized video image using both an optimal band-split filter phase construction function or a discrete wavelet transform 9/7 bi-orthogonal sub-band wavelet function; and

selecting the function having the best coding efficiency at a common quantization level.

31. The method of claim 30, further including applying the selected function to generate at least one signal-to-noise layer from the digitized video image.

32. A method for compressing a sequence of digitized video images represented as pixels, the method including:

applying a filter structure to generate a hierarchy of reduced resolution layers from a digitized video image; and

generating at least one signal-to-noise layer, each such signal-to-noise layer corresponding to one of the reduced resolution layers.

33. A method for compressing a sequence of digitized video image frames represented as pixels, the method including:

generating at least one processed layer from at least one of the digitized video image frames, each processed layer comprising a plurality of values; and

encoding at least one processed layer by preferentially allocating coding bits to regions of interest in such layer by decreasing the amplitude of pixel differences near edges of a digitized video image frame using a weighting which decreases as a function of nearness to such frame edges.

34. The method of claim 33, wherein at least one layer is a band-split layer layer.

35. The method of claim 33, wherein at least one layer is a signal-to-noise layer.

36. A method of using signal-to-noise improvement layers to improve compression of a sequence of images represented as pixels having a values representing different colors within a color space, the method including:

applying a non-linear function to each color value of the pixels of an image to create a perceptually weighted image; and

generating a signal-to-noise layer from each perceptually weighted image.

37. The method of claim 36, wherein the non-linear function is a gamma function.

39. A method for compressing a sequence of digitized video images in a motion compensated compression system, the method including:

generating a band-split hierarchy from a digitized video image, the band-split hierarchy comprising at least two different reduced resolution levels;

subtracting a previously decoded video image corresponding to a reduced resolution level from such reduced resolution level to create an intermediate frame;

generating a band-split hierarchy from the intermediate frame;

quantizing and dequantizing all levels, except the lowest level, of the intermediate frame into corresponding lower resolution images;

combined such lower resolution images and the lowest level of the intermediate frame into increasingly higher resolution coded images; and

adding the highest resolution coded image to the previously decoded video image to generate a viewable image.

40. A method for compressing a sequence of digitized video images, the method including:

generating a band-split hierarchy from a digitized video image, the band-split hierarchy comprising at least two different reduced resolution levels;

determining a quantization scale factor as the extreme pixel value from a selected reduced resolution level;

applying the determined quantization scale factor as a quantization parameter for all reduced resolution levels having a higher resolution than the selected reduced resolution level.

41. The method of claim 40, further including selecting a specific pixel value from the selected reduced resolution level as the quantization scale factor.

42. The method of claim 41, further including determining a plurality of quantization scale factors, each corresponding to a reduced resolution level.

43. The method of claim 41, further including determining the extremum pixel value as a function of the values of one or more adjacent pixels.

44. A method for generating a signal-to-noise correction layer from a digitized video image represented as pixels, the method including:

quantizing and dequantizing the digitized video image to create a coded image;

subtracting the coded image from the digitized video image to create a signal-to-noise difference image represented as pixels; and

applying one or both of quantization weighting and pixel weighting to the signal-to-noise difference image pixel values such that larger difference image pixel values have smaller quantized values.

45. The method of claim 44, further including constraining each signal-to-noise difference image pixel value as a function of one or more adjacent pixels.

46. The method of claim 45, further including applying an additional quantization weighting as a function of the amount that an error value for a signal-to-noise difference image pixel value differs from an average error value for the signal-to-noise difference image.

47. A method for compressing a sequence of digitized video images represented as pixels, the method including:

generating a plurality of processed layers from at least one of the digitized video images; and

applying a corresponding quantization parameter to each processed layer, each corresponding quantization parameter being scaled such that lower resolution layers have lower quantization parameter values relative to higher resolution layers.

48. The method of claim 47, further including varying the value of the quantization parameter for each processed layer as a function of region within such processed layer.

49. A method for reproducing the appearance of film grain or camera noise in compressed digitized video images, the method including:

applying a transform amplitude histogram as a function of one or more of color, brightness, and region for at least one signal-to-noise layer generated from such compressed digitized video images.