在上一篇文章中,我们根据测出的四组psnr与bit rate之间的关系,根据这个关系以及VCEG-M34的提议拟合出相对的曲线。
现在需要计算你所采用的方法对psnr和bit rate 的影响。那么就需要求对应相同的psnr时候,码流的变化,和对应相同的码流是对应的psnr的变化。对于后者,我们根据上篇文章拟合出来的公式即可获得。
对于前者,我们需要知道其反函数才能求出对应的值。
然而,对于符号表达的参数如何求其反函数呢?matalb提供了相应的解决方法。首先先需要下面这三个函数。
syms: 指定符号变量,如,syms x,t,y
sym(finverse(myfun(par0,para1)));
subs( )
下面通过举例来求得反函数计算:
%
fmyfun.m
function y = fmyfun(para,snr)
syms bit
f = sym(finverse(myfun(para,bit))) ;
y = subs(f, ' bit ' ,snr);
function y = fmyfun(para,snr)
syms bit
f = sym(finverse(myfun(para,bit))) ;
y = subs(f, ' bit ' ,snr);
这样,就可以实现在已知snr的情况下,求得bit rate
下面给出一个应用例子:
SNR_org
=
[
% blue_sky.yuv
47.658 , 45.278 , 43.129 , 40.932 ;
% pedestrian_area.yuv
47.362 , 44.574 , 42.903 , 41.470 ;
% riverbed.yuv
46.845 , 43.768 , 41.320 , 39.330 ;
% rush_hour.yuv
47.000 , 44.486 , 43.223 , 42.086 ;
% station2.yuv
46.700 , 43.697 , 41.978 , 40.400 ;
% sunflower.yuv
47.156 , 45.147 , 43.680 , 41.922 ;
% tractor.yuv
46.903 , 43.875 , 41.375 , 39.215 ;
];
bit_org = [
% blue_sky.yuv
46552.18 , 19108.14 , 8675.51 , 4435.90 ;
% pedestrain_area.yuv
62321.42 , 24765.34 , 9946.34 , 5424.77 ;
% river_bed
125062.30 , 75294.25 , 44999.84 , 27501.96 ;
% rush_hour
60778.31 , 21838.84 , 8316.55 , 4418.00 ;
% station2
67541.70 , 20353.78 , 4219.07 , 1900.40 ;
% sunflower
40896.95 , 12166.38 , 5558.72 , 2951.06 ;
% tractor
97024.69 , 49946.76 , 23004.60 , 11202.05 ;
];
bit_org = bit_org / 1000 ;
SNR_25candi_us = [
% blue_sky
47.655 , 45.270 , 43.122 , 40.918 ;
% pedestrain_area
47.356 , 44.569 , 42.902 , 41.467 ;
% riverbed
46.847 , 43.764 , 41.316 , 39.323 ;
% rush hour
46.997 , 44.479 , 43.212 , 42.076 ;
% station2
46.690 , 43.688 , 41.965 , 40.388 ;
% sunflower
47.142 , 45.130 , 43.670 , 41.917 ;
% tractor
46.894 , 43.855 , 41.355 , 39.198 ;
];
bit_25candi_us = [
% blue_sky
46785.35 , 19217.93 , 8763.00 , 4475.66
% pedestrain_area
62328.03 , 24823.76 , 9958.28 , 5435.45 ;
% riverbed
124661.84 , 75122.54 , 44870.01 , 27449.57 ;
% rush_hour
60824.93 , 21945.10 , 8330.94 , 4427.84 ;
% station2
67662.64 , 20556.07 , 4223.99 , 1892.38 ;
% sunflower
41094.02 , 12226.13 , 5552.23 , 2954.66 ;
% tractor
97280.39 , 50153.28 , 23102.80 , 11236.82 ;
];
bit_25candi_us = bit_25candi_us / 1000 ;
av_snr = zeros( 1 , 7 );
av_bit = zeros( 1 , 7 );
for seq = 1 : 1 : 7
bit_T8x8off = bit_25candi_us(seq,:);
snr_T8x8off = SNR_25candi_us(seq,:);
[para_T8x8off,res] = lsqcurvefit(@myfun,ones( 1 , 4 ),bit_T8x8off,snr_T8x8off);
bit_T8x8on = bit_org(seq,:);
snr_T8x8on = SNR_org(seq,:);
[para_T8x8on,res] = lsqcurvefit(@myfun,ones( 1 , 4 ),bit_T8x8on,snr_T8x8on);
bit_min = max(bit_T8x8off( 1 ),bit_T8x8on( 1 ));
bit_max = min(bit_T8x8off( 4 ),bit_T8x8on( 4 ));
bit = bit_max: 0.1 :bit_min;
snr_off = myfun(para_T8x8off,bit);
snr_on = myfun(para_T8x8on, bit);
snr_min = max(snr_T8x8off( 1 ),snr_T8x8on( 1 ));
snr_max = min(snr_T8x8off( 4 ),snr_T8x8on( 4 ));
snr = snr_max: 0.1 :snr_min;
bit_off = fmyfun(para_T8x8off,snr);
bit_on = fmyfun(para_T8x8on, snr);
er_snr = snr_on - snr_off;
er_bit = (bit_on - bit_off). / bit_on;
av_snr( 1 ,seq) = sum(er_snr) / (size(er_snr, 2 ));
av_bit( 1 ,seq) = sum(er_bit) / (size(er_bit, 2 ));
% blue_sky.yuv
47.658 , 45.278 , 43.129 , 40.932 ;
% pedestrian_area.yuv
47.362 , 44.574 , 42.903 , 41.470 ;
% riverbed.yuv
46.845 , 43.768 , 41.320 , 39.330 ;
% rush_hour.yuv
47.000 , 44.486 , 43.223 , 42.086 ;
% station2.yuv
46.700 , 43.697 , 41.978 , 40.400 ;
% sunflower.yuv
47.156 , 45.147 , 43.680 , 41.922 ;
% tractor.yuv
46.903 , 43.875 , 41.375 , 39.215 ;
];
bit_org = [
% blue_sky.yuv
46552.18 , 19108.14 , 8675.51 , 4435.90 ;
% pedestrain_area.yuv
62321.42 , 24765.34 , 9946.34 , 5424.77 ;
% river_bed
125062.30 , 75294.25 , 44999.84 , 27501.96 ;
% rush_hour
60778.31 , 21838.84 , 8316.55 , 4418.00 ;
% station2
67541.70 , 20353.78 , 4219.07 , 1900.40 ;
% sunflower
40896.95 , 12166.38 , 5558.72 , 2951.06 ;
% tractor
97024.69 , 49946.76 , 23004.60 , 11202.05 ;
];
bit_org = bit_org / 1000 ;
SNR_25candi_us = [
% blue_sky
47.655 , 45.270 , 43.122 , 40.918 ;
% pedestrain_area
47.356 , 44.569 , 42.902 , 41.467 ;
% riverbed
46.847 , 43.764 , 41.316 , 39.323 ;
% rush hour
46.997 , 44.479 , 43.212 , 42.076 ;
% station2
46.690 , 43.688 , 41.965 , 40.388 ;
% sunflower
47.142 , 45.130 , 43.670 , 41.917 ;
% tractor
46.894 , 43.855 , 41.355 , 39.198 ;
];
bit_25candi_us = [
% blue_sky
46785.35 , 19217.93 , 8763.00 , 4475.66
% pedestrain_area
62328.03 , 24823.76 , 9958.28 , 5435.45 ;
% riverbed
124661.84 , 75122.54 , 44870.01 , 27449.57 ;
% rush_hour
60824.93 , 21945.10 , 8330.94 , 4427.84 ;
% station2
67662.64 , 20556.07 , 4223.99 , 1892.38 ;
% sunflower
41094.02 , 12226.13 , 5552.23 , 2954.66 ;
% tractor
97280.39 , 50153.28 , 23102.80 , 11236.82 ;
];
bit_25candi_us = bit_25candi_us / 1000 ;
av_snr = zeros( 1 , 7 );
av_bit = zeros( 1 , 7 );
for seq = 1 : 1 : 7
bit_T8x8off = bit_25candi_us(seq,:);
snr_T8x8off = SNR_25candi_us(seq,:);
[para_T8x8off,res] = lsqcurvefit(@myfun,ones( 1 , 4 ),bit_T8x8off,snr_T8x8off);
bit_T8x8on = bit_org(seq,:);
snr_T8x8on = SNR_org(seq,:);
[para_T8x8on,res] = lsqcurvefit(@myfun,ones( 1 , 4 ),bit_T8x8on,snr_T8x8on);
bit_min = max(bit_T8x8off( 1 ),bit_T8x8on( 1 ));
bit_max = min(bit_T8x8off( 4 ),bit_T8x8on( 4 ));
bit = bit_max: 0.1 :bit_min;
snr_off = myfun(para_T8x8off,bit);
snr_on = myfun(para_T8x8on, bit);
snr_min = max(snr_T8x8off( 1 ),snr_T8x8on( 1 ));
snr_max = min(snr_T8x8off( 4 ),snr_T8x8on( 4 ));
snr = snr_max: 0.1 :snr_min;
bit_off = fmyfun(para_T8x8off,snr);
bit_on = fmyfun(para_T8x8on, snr);
er_snr = snr_on - snr_off;
er_bit = (bit_on - bit_off). / bit_on;
av_snr( 1 ,seq) = sum(er_snr) / (size(er_snr, 2 ));
av_bit( 1 ,seq) = sum(er_bit) / (size(er_bit, 2 ));
end