Method and system for sampling video data
View Patent ↗A method and system for sampling video data uses re-sampling filters having lengths optimized relative to a quantization parameter of video processing. The method uses modeling of an optimal length of the re-sampling filter as a function of the quantization parameter to derive empirical formulas and a look up table for optimal lengths of re-sampling filters. The resulting re-sampling filters are selectively adapted for sampling video data having different bit rates.
1. A method comprising:
(a) modeling, by a data processing system, an optimal length (L OPT ) of a re-sampling filter as a function of a quantization parameter (QP) of the re-sampling filter;
(b) deriving, by the data processing system, empirical formulas for the optimal length using results of the step (a);
(c) producing, by the data processing system, a look up table for the optimal length using at least one of the empirical formulas;
(d) designing, by the data processing system, a re-sampling filter having the optimal length defined using the empirical formulas or the look up table; and
(e) sampling video data using the re-sampling filter.
2. The method of claim 1 , wherein the step (a) further comprises:
defining the quantization parameter (QP) as a measure of a compression ratio of the re-sampling filter;
expressing a value of the quantization parameter (QP) in units of the compression ratio;
expressing the optimal length (L OPT ) in units of taps of the re-sampling filter; and
defining the optimal length (L OPT ) as a minimal length providing that a peak signal-to-noise ratio (PSNR) of sampled video data is equal to or exceeds a pre-determined limit.
3. The method of claim 1 , wherein the step (a) further comprises:
determining coding efficiency of the re-sampling filter.
4. The method of claim 1 , wherein the step (b) further comprises:
deriving the empirical formulas where the optimal length L OPT is a function L OPT (QP) of the quantization parameter QP.
5. The method of claim 4 , further comprising:
deriving the empirical formulas where the optimal length L OPT is at least one of a linear function, an exponential function, or a polynomial function of the quantization parameter QP.
6. The method of claim 5 , wherein the linear function is:
L OPT ( QP )=1.0757( QP )+2.1205, with a statistic R -squared value of about 0.8734.
7. The method of claim 5 , wherein the exponential function is:
L OPT ( QP )=3.1397e 0.1575(QP) , with a statistic R -squared value of about 0.9115.
8. The method of claim 5 , wherein the polynomial function is:
L OPT ( QP )=L OPT ( QP )=0.1561( QP ) 2 −0.2508( QP )+3.9629, with a statistic R -squared value of about 0.9653.
9. The method of claim 1 , wherein the step (c) further comprises:
producing the look up table where the optimal length (L OPT ) of the re-sampling filter is expressed as a function of the quantization parameter (QP) as follows:
QUANTIZATION PARAMETER (QP)
OPTIMAL LENGTH (L OPT )
0-12
4
21-32
6
33-38
8
39-44
10
44-51
12
10. The method of claim 1 , wherein the step (d) further comprises:
determining the quantization parameter (QP) and a peak signal-to-noise ratio (PSNR) of the re-sampling filter.
11. The method of claim 1 , wherein the re-sampling filter is a low-pass digital filter.
12. A non-transitory computer readable medium storing program code that, when executed by a data processor, facilitates execution of the method of claim 1 .
13. A system for sampling video data, comprising:
a delay module for delaying the video data;
a bit rate analyzer for determining a quantization parameter (QP) for a re-sampling filter;
a reference module for determining an optimal length (L OPT ) of the re-sampling filter based on the QP;
the re-sampling filter adapted to the optimal length (L OPT ) based on a bit rate of the video data; and
a processor, responsive to the reference module, adapted to determine transform coefficients of the re-sampling filter.
14. The system of claim 13 , wherein at least one of the delay module, the bit rate analyzer, the reference module, or the re-sampling filter is implemented programmable or non-programmable hardware, or a combination thereof.
15. The system of claim 13 , wherein:
the quantization parameter (QP) is a measure of a compression ratio; and
the optimal length (L OPT ) is a minimal length providing that a peak signal-to-noise ratio (PSNR) of the sampled video data is equal to or exceeds a pre-determined limit.
16. The system of claim 13 , wherein the reference module uses at least one of empirical formulas A, B, or C to calculate the optimal length L OPT expressed as a function L OPT (QP) of the quantization parameter QP:
L OPT ( QP )=1.0757( QP )+2.1205, with a statistic R -squared value of about 0.8734; (A)
L OPT ( QP )=3.1397e 0.1575(QP) , with a statistic R -squared value of about 0.9115; and (B)
L OPT ( QP )= L OPT ( QP )=0.1561( QP ) 2 −0.2508( QP )+3.9629, with a statistic R -squared value of about 0.9653. (C)
17. The system of claim 13 , wherein the reference module contains a look up table where the optimal length (L OPT ) of the re-sampling filter is expressed as a function of the quantization parameter (QP) as follows:
QUANTIZATION PARAMETER (QP)
OPTIMAL LENGTH (L OPT )
0-12
4
21-32
6
33-38
8
39-44
10
44-51
12
18. The system of claim 13 , further comprising:
a plurality of re-sampling filters, wherein each re-sampling filter has the optimal length (L OPT ) for sampling video data having a particular bit rate; and
a selector directing the video data to the re-sampling filter having the optimal length (L OPT ) for sampling the video data.
19. The system of claim 18 , wherein the selector de-multiplexes time-multiplexed video data having different bit rates and selectively directs the de-multiplexed video data to the re-sampling filters having the optimal lengths (L OPT ) for sampling the respective portion of the de-multiplexed video data.
20. The system of claim 13 , wherein the re-sampling filter is a low-pass digital filter.