aviad rozenhek wrote:
>> is there a clear way to put the swscale flags on a quality scale from low
>>     
> to high?
>   
>> i.e. is it true that in terms of quality and cpu usage:
>> SWS_FAST_BILINEAR < SWS_BILINEAR < SWS_BICUBIC < SWS_X < SWS_POINT <
>>     
> SWS_AREA < SWS_BICUBLIN < SWS_GAUSS < SWS_SINC < SWS_LANCZOS < SWS_SPLINE ?
>
> I wrote a little program to test this.
> I tried to order them from fast to slow and from crude to fine quality on a
> single scale.
> I found out that I have to remove SWS_AREA, SWS_X, SWS_GAUSS, SWS_LANCZOS
>
>
> the scale is:
>
> FAST/CRUDE
>              SLOW/FINE
> --------------------------------------------------------------------------------------------------
> SWS_POINT <  SWS_FAST_BILINEAR < SWS_BILINEAR < SWS_BICUBLIN < SWS_BICUBIC
> < SWS_SPLINE < SWS_SINC
>                                  SWS_AREA
>  SWS_LANCZOS
>                                  SWS_X
>                                  SWS_GAUSS
>
> here is some raw data, gathered by downsampling 640x480 to 320x240 and back
> again to 640x480, using mmx+mmx2 on E6400 intel
>
> errors(measured in sum-squared-differences)
> SWS_POINT   SWS_FAST_BILINEAR  SWS_BILINEAR  SWS_BICUBLIN  SWS_BICUBIC
> SWS_SPLINE  SWS_SINC
> 3769328057  2636583475         2155004550    2039237301    1980996866
>  1951467852  1902084412
>
> time(how much time to downsample, copy and upsample, in seconds)
> SWS_POINT  SWS_FAST_BILINEAR   SWS_BILINEAR  SWS_BICUBLIN  SWS_BICUBIC
>  SWS_SPLINE  SWS_SINC
> 1.1034     1.2289              1.7151        2.2899        2.6026
> 6.719       13.2753
>
>
> I hope someone finds it useful. if you want the code i'd be happy to
> contribute.
> Aviad
>   


Hello Aviad,

These are results in seconds for ONE iteration.  It's unbelieavably slow!!!!
Can you please post the code?


Thanks
Vadim

_______________________________________________
libav-user mailing list
[email protected]
https://lists.mplayerhq.hu/mailman/listinfo/libav-user

Reply via email to