[x265] Compression gains...results update

[Mario Rohkrämer]

Remember, x265 optimizes for psycho-visually convenient results per  
default, which may look better to many, but may have worse PSNR and SSIM  
values.

If you want to compare SSIM values with another encoder, use "--tune  
ssim"; if you want to compare PSNR results with another encoder, use  
"--tune psnr". Optimizing for more than one metric at the same time does  
not make sense.


Am 29.07.2016, 14:55 Uhr, schrieb N Vijay Anand  
<nvijay.anand at trispacetech.com>:

> Regression still on...
>
> --bitrate xxx --psnr --ssim --me hex --input-res 1920x1080 --fps 30
>
> psnr is Global PSNR:
> ssim is SIM Mean Y:
>
> Bitrate(Kbps) x265 Tri
> psnr ssim psnr ssim
> 64 32.369000 0.822945 31.921000 0.810605
> 128 32.393000 0.824781 32.055000 0.818082
> 192 32.508000 0.831349 32.042000 0.819762
> 256 33.235000 0.846821 32.895000 0.835494
> 320 34.144000 0.860380 33.806000 0.849388
> 384 34.890000 0.872120 34.502000 0.860695
> 448 35.465000 0.881114 35.037000 0.869120
> 512 36.020000 0.890695 35.475000 0.876691
> 576 36.521000 0.899405 35.651000 0.879254
> 640 36.963000 0.906685 36.391000 0.892394
> 704 37.370000 0.913123 36.558000 0.895628
> 768 37.730000 0.918413 36.952000 0.901972
> 832 38.038000 0.922564 37.215000 0.906528
> 896 38.336000 0.926395 37.470000 0.910345
> 960 38.603000 0.929653 37.668000 0.913329
> 1024 38.832000 0.932241 37.983000 0.917706
> 2048 41.108000 0.953156 40.411000 0.943970
> 3072 42.257000 0.960902 41.669000 0.954744
> 4096 43.027000 0.966252 42.580000 0.961890
> 5120 43.562000 0.969705 43.093000 0.965873
> 6144 43.934000 0.971961 43.642000 0.969464
> 7168 44.212000 0.973557 43.901000 0.971225
> 8192 44.430000 0.974746 44.204000 0.973049
>
>
> when the requirement is specific target bit-rate, x265 is better.
> But if requirement is "generic",  you need take call on compression  
> gains.
>


-- 

Fun and success!
Mario *LigH* Rohkrämer
mailto:contact at ligh.de