Google's New VP8 Codec SDK Is Better, Faster
Following Google opening up the VP8 video codec specification in May and launching the WebM container format, in July the developers behind FFmpeg created the ffvp8 decoder that was much faster than Google's own VP8 decoding library. Google has now, however, provided a new version of the VP8 Codec SDK that they have codenamed "Aylesbury" and it's designed to be better and much faster than their original release.
Compared to the 19 May libvpx state, Aylesbury is 20~40% faster with its decoding speed and there's a 7% overall PSNR (peak signal-to-noise) improvement with VP8's best quality encoding mode. There's also up to a 60% improvement on very noisy, still or slow moving source videos, according to Google engineers.
Making some of these improvements possible are single-core assembly optimizations, threading improvements for multi-core CPUs, improved memory handling, a smaller memory foot-print, combining IDCT and reconstruction operations, and SSE3 usage where supported.
The next VP8 Codec SDK release from Google is expected in Q1'2011 and is codenamed Bali. Details of the Aylesbury release can be found on the WebM blog.
Compared to the 19 May libvpx state, Aylesbury is 20~40% faster with its decoding speed and there's a 7% overall PSNR (peak signal-to-noise) improvement with VP8's best quality encoding mode. There's also up to a 60% improvement on very noisy, still or slow moving source videos, according to Google engineers.
Making some of these improvements possible are single-core assembly optimizations, threading improvements for multi-core CPUs, improved memory handling, a smaller memory foot-print, combining IDCT and reconstruction operations, and SSE3 usage where supported.
The next VP8 Codec SDK release from Google is expected in Q1'2011 and is codenamed Bali. Details of the Aylesbury release can be found on the WebM blog.
25 Comments