Digital image coding system having self-adjusting selection criteria for selecting a transform function

• US 20040032988A1
• Filed: 08/07/2003
• Published: 02/19/2004
• Est. Priority Date: 07/03/1996
• Status: Active Grant

• Current Assignee: Apple Inc.
• Original Assignee: Apple Inc.
• Inventors: Lu, Jian, Wu, Hsi-Jung, Chu, Ke-Chiang, Tian, Yu Tina
• Granted Patent: US 7,079,695 B2
• Time in Patent Office: Days
• US Class Current: 382/239
• CPC Class Codes: G06T 9/007 Transform coding, e.g. discrete cosine transform
• CPC Class Codes: H04N 19/122 Selection of transform size, e.g. 8x8 or 2x4x8 DCT; Selection of sub-band transforms of varying structure or type
• CPC Class Codes: H04N 19/132 Sampling, masking or truncation of coding units, e.g. adaptive resampling, frame skipping, frame interpolation or high-frequency transform coefficient masking
• CPC Class Codes: H04N 19/139 Analysis of motion vectors, e.g. their magnitude, direction, variance or reliability
• CPC Class Codes: H04N 19/14 Coding unit complexity, e.g. amount of activity or edge presence estimation H04N19/146 takes precedence
• CPC Class Codes: H04N 19/154 Measured or subjectively estimated visual quality after decoding, e.g. measurement of distortion use of rate-distortion criteria H04N19/147
• CPC Class Codes: H04N 19/176 the region being a block, e.g. a macroblock
• CPC Class Codes: H04N 19/18 the unit being a set of transform coefficients
• CPC Class Codes: H04N 19/196 being specially adapted for the computation of encoding parameters, e.g. by averaging previously computed encoding parameters processing of motion vectors H04N19/513
• CPC Class Codes: H04N 19/198 including smoothing of a sequence of encoding parameters, e.g. by averaging, by choice of the maximum, minimum or median value
• CPC Class Codes: H04N 19/61 in combination with predictive coding

Abstract

In a digital signal processing system, a method for selecting a transform function to apply to an input signal based on characteristics of the signal, and for self-adjusting criteria which are used in selecting a transform function to apply to a subsequent signal. Characteristics are obtained from the signal. The characteristics are compared to adjustable criteria which are used in selecting a transform function. Differing criteria are maintained for the different selectable transform functions. A record is maintained of transform functions selected and the particular characteristics that caused the selection. Based on the ability of a transform function to minimally define the coded signal, an inverse transform function is selected to decode the signal. The criteria used in selecting a transform function to apply to a subsequent signal are adjusted based on a quality measure of the decoded signal and the record of selected transform functions.

Claims

1. A computer-implemented method for coding a block of pixels of a digitized video image using a selectable one of a plurality of coding functions, comprising the steps of:

establishing adjustable selection criteria for selecting a coding function;

measuring a predetermined characteristic of the block to obtain a characteristic value;

selecting a coding function based on said selection criteria and said characteristic value;

coding the block according to said coding function from said selecting step to obtain a coded block;

performing a quality measurement of said coded block; and

adjusting said selection criteria based on said quality measurement, whereby quality measures of subsequent blocks are improved through adjusted selection criteria for selecting coding functions.

2. The method of claim 1, wherein said performing a quality measurement step comprises the steps of:

decoding said coded block to obtain a decoded block; and

performing a qualiy measurement of said decoded block.

3. The method of claim 2, further comprising the step of selecting a decoding function independent of selecting said coding function.

4. The method of claim 3, further comprising the steps of:

accumulating an historical record of detected characteristics of blocks and selected decoding functions; and

periodically adjusting said selection criteria based on said historical record.

5. The method of claim 1, wherein said measuring step measures a predetermined plurality of characteristics, and said selecting step selects a coding function based on said selection criteria and said plurality of characteristics.

6. The method of claim 5, wherein said plurality of characteristics include a total energy characteristic defined as Σ

<sub>(i,j)∈

block</sub>|x(i, j)|, wherein x represents the block of pixels, and i and j are indices into the block.

7. The method of claim 5, wherein said plurality of characteristics include a horizontal high pass energy characteristic defined as

Σ

<sub>0≦

i<

Blockwidth−

1, 0≦

j<

BlockHeight</sub>|x(i, j)−

x(i+1,j)|

wherein x represents the block of pixels, and i and j form indices into the block.

8. The method of claim 5, wherein said plurality of characteristics include a vertical high pass energy characteristic defined as

Σ

<sub>0≦

i<

Blockwidth, 0≦

j<

BlockHeight−

1</sub>|x(i, j)−

x(i,j+1)|

wherein x represents the block of pixels, and i and j form indices into the block.

9. The method of claim 5, wherein said plurality of characteristics include a motion vector magnitude characteristic, defined as x²+y², wherein x represents the movement of the image in the block on an x-axis and y represents movement of the image in the block on a y-axis.

10. The method of claim 5, wherein said plurality of characteristics include a motion vector magnitude characteristic defined as the maximum of x and y, wherein x represents the movement of the image in the block on an x-axis and y represents movement of the image in the block on a y-axis.

11. The method of claim 1, wherein said step of performing a quality measurement includes the step of obtaining a peak-signal-to-noise ratio.

12. The method of claim 11, further comprising the steps of:

accumulating an historical record of detected characteristics of blocks and selected decoding functions; and

periodically adjusting said selection criteria based on said historical record.

13. A computer-implemented method for transforming a block of pixels within a frame of a digitized video image using a selectable one of a set of transform functions, each transform function having an inverse transform function, and each block having a predetermined set of image characteristics, the method comprising the steps of:

establishing adjustable selection criteria for selecting a transform function;

detecting respective characteristic values for the image characteristics of a block;

selecting a first transform function from the set of transform functions based on said characteristic values and said selection criteria;

applying said first transform function to the block to form a transformed block;

quantizing said transformed block to form a quantized block;

selecting an inverse transform function whose application minimally covers said quantized block;

inversely quantizing said quantized block to form an inversely quantized block;

applying said inverse transform function to said inversely quantized block to form a decoded block;

establishing a quality value for said decoded block; and

updating said selection criteria based on said quality value and said characteristic values.

14. A computer-implemented method for transforming a block of pixels within a frame of a digitized video image using a selectable one of a set of transform functions, each transform function having an inverse transform function, and each block having a predetermined set of image characteristics, the method comprising the steps of:

associating the transform functions with the set of image characteristics, with predetermined quantization values, and with adjustable thresholds associated with the image characteristics;

obtaining respective characteristic values for the image characteristics of a block;

selecting a transform function from the set of transform functions based on comparisons between said characteristic values and said adjustable thresholds;

applying said transform function to the block to form a transformed block;

quantizing said transformed block using a quantization value to form a quantized block;

selecting an inverse transform function whose application minimally covers said quantized block;

inversely quantizing said quantized block to form an inversely quantized block;

applying said inverse transform function to said inversely quantized block to form a decoded block;

obtaining a quality value for said decoded block; and

updating said adjustable thresholds based on said quality value and said characteristic values.

15. The method of claim 14, further comprising the steps of:

establishing histograms of characteristic values for each of the image characteristics and associated transform functions and quantization values;

recording said characteristic values in histograms referenced by said transform function and said quantization value;

selecting a statistical function to apply to said histograms;

applying said statistical function to said histograms; and

updating said adjustable thresholds with data from application of said statistical function to said histograms.

16. The method of claim 14, further comprising the steps of:

establishing histograms of quality values and characteristic values and associated transform functions and quantization values;

recording said quality value and characteristic values in histograms referenced by said transform function and said quantization value;

selecting a statistical function to apply to said histograms;

applying said statistical function to said histograms; and

updating said adjustable thresholds with data from application of said statistical function to said histograms.

17. The method of claim 14, wherein said step of obtaining characteristic values comprises the steps of:

obtaining a total energy value;

obtaining a horizontal high pass energy value;

obtaining a vertical high pass energy value; and

obtaining a motion vector magnitude value.

18. The method of claim 17, wherein said step of selecting a transform function comprises the steps of:

comparing said total energy value to a first threshold;

comparing said horizontal high pass energy value to a second threshold;

comparing said vertical high pass energy value to a third threshold;

comparing said motion vector magnitude value to a fourth threshold; and

if for a given transform function, said total energy value is less than said first threshold, and said horizontal high pass energy value is less then a second threshold, and said vertical high pass energy value is less than a third threshold, and said motion vector magnitude value is less than a fourth threshold, then selecting said given transform function.

19. The method of claim 17, wherein said establishing histograms step further comprises the steps of:

establishing a total energy histogram;

establishing a horizontal high pass energy histogram;

establishing a vertical high pass energy histogram; and

establishing a motion vector magnitude histogram.

20. The method of claim 19, wherein said step of selecting a transform function comprises the steps of:

comparing said total energy value to a first threshold;

comparing said horizontal high pass energy value to a second threshold;

comparing said vertical high pass energy value to a third threshold;

comparing said motion vector magnitude value to a fourth threshold; and

if for a given transform function, said total energy value is less than said first threshold, and said horizontal high pass energy value is less then a second threshold, and said vertical high pass energy value is less than a third threshold, and said motion vector magnitude value is less than a fourth threshold, then selecting said given transform function.

21. The method of claim 15, wherein said step of obtaining a quality value further comprises the step of obtaining a peak signal-to-noise ratio for said decoded block.

22. The method of claim 21, wherein said step of updating said adjustable thresholds further comprises the steps of:

selecting an order statistic based on the quality value, transform function, and quantizer value;

applying said order statistic to said histograms to obtain new thresholds; and

updating said adjustable thresholds with said new thresholds.

23. A processor for coding a block of pixels of a digitized video image using a selectable one of a plurality of coding functions, comprising:

establishment means for establishing adjustable selection criteria used to select a coding function;

measurement means for measuring a predetermined characteristic of the block to obtain a characteristic value;

selection means for selecting a coding function based on said selection criteria and said characteristic value;

code means for coding the block according to said coding function from said selecting step to obtain a coded block;

performance means for performing a quality measurement of said coded block; and

adjustment means for adjusting said selection criteria based on said quality measurement, whereby quality measures of subsequent blocks are improved through adjusted selection criteria for selecting coding functions.

24. A computer-readable medium comprising program instructions for causing a computer to code a block of pixels of a digitized video image using a selectable one of a plurality of coding functions, by performing the steps of:

establishing adjustable selection criteria for selecting a coding function;

measuring a predetermined characteristic of the block to obtain a characteristic value;

selecting a coding function based on said selection criteria and said characteristic value;

coding the block according to said coding function from said selecting step to obtain a coded block;

performing a quality measurement of said coded block; and

adjusting said selection criteria based on said quality measurement, whereby quality measures of subsequent blocks are improved through adjusted selection criteria for selecting coding functions.