Low cost rate-distortion computations for video compression
A method of rate-distortion computations for video compression is disclosed. The method may include steps (A) to (C). Step (A) may generate a plurality of transform coefficients from a residual block of the video using a circuit. Step (B) may generate a block distortion value (i) based on the transform coefficients and (ii) independent of a plurality of inverse transform samples produced from the residual block. Step (C) may generate a rate-distortion value from the block distortion value.
1. A method for estimating a number of bits used to code an image block, comprising the steps of:
transforming said image block in a circuit to calculate a plurality of transform coefficients;
quantizing said transform coefficients to calculate a plurality of quantized coefficients;
counting non-zero ones of said quantized coefficients;
translating said count of said non-zero quantized coefficients using a curve stored in a lookup table to produce a rate value that estimates said number of bits used to code said image block;
generating a plurality of intermediate coefficients by calculating absolute values of said transform coefficients; and
generating a block distortion value based on a difference between scaled versions of said quantized coefficients and said intermediate coefficients.
2. The method according to claim 1 , wherein said generation of said block distortion value is independent of a plurality of inverse quantized coefficients corresponding to said quantized coefficients.
3. The method according to claim 1 , wherein said curve stored in said lookup table comprises a non-linear curve.
4. The method according to claim 1 , further comprising the steps of:
inverse quantizing said quantized coefficients to calculate a plurality of inverse quantized coefficients; and
calculating a first difference between a first one of said transform coefficients and a first one of said inverse quantized coefficients.
5. The method according to claim 4 , further comprising the step of:
calculating a second difference between a second one of said transform coefficients and a second one of said inverse quantized coefficients.
6. The method according to claim 1 , further comprising the step of:
generating a score of said image block based on said rate value and said block distortion value.
7. An apparatus comprising:
a first circuit configured to transform an image block to calculate a plurality of transform coefficients; and
a second circuit configured to
(i) quantize said transform coefficients to calculate a plurality of quantized coefficients,
(ii) count non-zero ones of said quantized coefficients,
(iii) translate said count of said non-zero quantized coefficients using a curve stored in a lookup table to produce a rate value that estimates a number of bits used to code said image block,
(iv) generate a plurality of intermediate coefficients by calculating absolute values of said transform coefficients, and
(v) generate a block distortion value based on a difference between scaled versions of said quantized coefficients and said intermediate coefficients.
8. The apparatus according to claim 7 , wherein said generation of said block distortion value is independent of a plurality of inverse quantized coefficients corresponding to said quantized coefficients.
9. The apparatus according to claim 7 , wherein said curve stored in said lookup table comprises a non-linear curve.
10. The apparatus according to claim 7 , wherein said second circuit is further configured to inverse quantize said quantized coefficients to calculate a plurality of inverse quantized coefficients, and calculate a first difference between a first one of said transform coefficients and a first one of said inverse quantized coefficients.
11. The apparatus according to claim 10 , wherein said second circuit is further configured to calculate a second difference between a second one of said transform coefficients and a second one of said inverse quantized coefficients.
12. The apparatus according to claim 7 , wherein said second circuit is further configured to generate a score of said image block based on said rate value and said block distortion value.
13. A method for estimating a number of bits used to code an image block, comprising the steps of:
transforming said image block in a circuit to calculate a plurality of transform coefficients;
quantizing said transform coefficients to calculate a plurality of quantized coefficients;
counting non-zero ones of said quantized coefficients;
translating said count of said non-zero quantized coefficients using a curve stored in a lookup table to produce a rate value that estimates said number of bits used to code said image block;
inverse quantizing said quantized coefficients to calculate a plurality of inverse quantized coefficients;
calculating a first difference between a first one of said transform coefficients and a first one of said inverse quantized coefficients;
calculating a second difference between a second one of said transform coefficients and a second one of said inverse quantized coefficients; and
generating a block distortion value of said image block based on a combined value of said first difference and said second difference, wherein said combined value is calculated as a weighted sum of squares of said first difference and said second difference.
14. The method according to claim 13 , further comprising the step of:
generating a plurality of intermediate coefficients by calculating absolute values of said transform coefficients.
15. The method according to claim 13 , wherein said curve stored in said lookup table comprises a non-linear curve.
16. The method according to claim 13 , further comprising the step of:
generating a score of said image block based on said rate value and said block distortion value.
17. An apparatus comprising:
a first circuit configured to transform an image block to calculate a plurality of transform coefficients; and
a second circuit configured to
(i) quantize said transform coefficients to calculate a plurality of quantized coefficients,
(ii) count non-zero ones of said quantized coefficients,—
(iii) translate said count of said non-zero quantized coefficients using a curve stored in a lookup table to produce a rate value that estimates a number of bits used to code said image block,
(iv) inverse quantize said quantized coefficients to calculate a plurality of inverse quantized coefficients,
(v) calculate a first difference between a first one of said transform coefficients and a first one of said inverse quantized coefficients,
(vi) calculate a second difference between a second one of said transform coefficients and a second one of said inverse quantized coefficients, and
(vii) generate a block distortion value of said image block based on a combined value of said first difference and said second difference, wherein said combined value is calculated as a weighted sum of squares of said first difference and said second difference.
18. The apparatus according to claim 17 , wherein said second circuit is further configured to generate a plurality of intermediate coefficients by calculating absolute values of said transform coefficients.
19. The apparatus according to claim 17 , wherein said curve stored in said lookup table comprises a non-linear curve.
20. The apparatus according to claim 17 , wherein said second circuit is further configured to generate a score of said image block based on said rate value and said block distortion value.