Low cost rate-distortion computations for video compression
An apparatus including a first circuit and a second circuit. The first circuit may be configured to transform an image block into a plurality of transform coefficients. The second circuit may be configured in each of a plurality of modes to (i) quantize the transform coefficients to calculate a plurality of quantized coefficients, (ii) translate a number of non-zero values of the quantized coefficients to a rate value, where the translation is a non-linear translation, (iii) calculate a distortion value (a) based on the transform coefficients and (b) independent of a plurality of inverse quantized coefficients corresponding to the quantized coefficients, and (iv) calculate a score of a current mode of the plurality of modes to encode the image block based on the rate value and the distortion value.
1. An apparatus comprising:
a first circuit configured to transform an image block into a plurality of transform coefficients; and
a second circuit configured in each of a plurality of modes to: (i) quantize said transform coefficients to calculate a plurality of quantized coefficients, (ii) translate a number of non-zero values of said quantized coefficients to a rate value, wherein said translation is a non-linear translation, (iii) calculate a distortion value (a) based on said transform coefficients and (b) independent of a plurality of inverse quantized coefficients corresponding to said quantized coefficients, and (iv) calculate a score of a current mode of said plurality of modes to encode said image block based on said rate value and said distortion value.
2. The apparatus according to claim 1 , further comprising a third circuit configured to select a best mode of said plurality of modes that corresponds a lowest score among said plurality of scores.
3. The apparatus according to claim 1 , wherein each of said modes comprises (i) one of an intra-prediction mode and an inter-prediction mode, (ii) one of a plurality of block sizes and (iii) one of a plurality of quantization parameters.
4. The apparatus according to claim wherein a slope at which said rate value increases is higher in said non-linear translation as said number of non-zero values of said quantized coefficients increases.
5. The apparatus according to claim 1 , wherein said second circuit is further configured to calculate a plurality of roundoff error values caused by said quantization of said transform coefficients.
6. The apparatus according to claim 5 , wherein said second circuit is configured to calculate said distortion value further based on said roundoff error values.
7. The apparatus according to claim 5 , wherein said roundoff error values are based on a quantization parameter used by said second circuit in said current mode.
8. The apparatus according to claim 5 , wherein said roundoff error values are based on a transform size used by said first circuit in said current mode.
9. The apparatus according to claim 1 , wherein said second circuit is further configured to generate a plurality of intermediate coefficients by calculating absolute values of said transform coefficients.
10. The apparatus according to claim 1 , wherein said distortion value is calculated based on a difference between scaled versions of said quantized coefficients and a plurality of absolute values of said transform coefficients.
11. A method for rate-distortion computations comprising the steps of:
transforming an image block into a plurality of transform coefficients; and
in each of a plurality of modes:
quantizing said transform coefficients to calculate a plurality of quantized coefficients;
translating a number of non-zero values of said quantized coefficients to a rate value, wherein said translation is a non-linear translation;
calculating a distortion value (a) based on said transform coefficients and (b) independent of a plurality of inverse quantized coefficients corresponding to said quantized coefficients; and
calculating a score of a current mode of said plurality of modes to encode said image block based on said rate value and said distortion value.
12. The method according to claim 11 , further comprising the step of:
selecting a best mode of said plurality of modes that corresponds a lowest score among said plurality of scores.
13. The method according to claim 11 , wherein each of said modes comprises (i) one of an intra-prediction mode and an inter-prediction mode, (ii) one of a plurality of block sizes and (iii) one of a plurality of quantization parameters.
14. The method according to claim 11 , wherein a slope at which said rate value increases is higher in said non-linear translation as said number of non-zero values of said quantized coefficients increases.
15. The method according to claim 11 , further comprising the step of:
calculating a plurality of roundoff error values caused by said quantization of said transform coefficients, respectively.
16. The method according to claim 15 , wherein said distortion value is calculated further based on said roundoff error values.
17. The method according to claim 15 , wherein said roundoff error values are based on a quantization parameter used by said quantization in said current mode.
18. The method according to claim 15 , wherein said roundoff error values are based on a transform size used by said transformation in said current mode.
19. The method according to claim 11 , further comprising the step of:
generating a plurality of intermediate coefficients by calculating absolute values of said transform coefficients.
20. The method according to claim 11 , wherein said distortion value is calculated based on a difference between scaled versions of said quantized coefficients and a plurality of absolute values of said transform coefficients.