Timothy Terriberry
http://lca2015.linux.org.au/schedule/30187/view_talk
With the recent completion of the next-generation video codecs VP9 and HEVC, Xiph has turned our attention to the "next-next generation". The Daala project aims to design a royalty-free codec that surpasses the performance of the best patent-encumbered formats.
Unlike the rest of the industry, we are not trying to achieve superior performance through dozens of minor optimizations of the same basic coding tools in use since the 1980's. Rather, we are trying some fundamentally different ideas. Some of these are based on old research directions that were never adequately explored. Some of these are new ideas inspired by our prior codec work. This strategy is riskier than making many small, incremental changes, but it offers us a chance to break out of the local minima of current codec design and helps avoid the dense patent thickets built up around the traditional approaches.
This talk will explain how Daala is different, why we think it will work, and how far we've come so far. It is intended for an audience with a technical background who does not already have significant signal processing experience.
