Extended dynamic texture prediction for H.264/AVC inter coding

This paper proposes an extension to an existing H.264/AVC encoder that improves the coding performance for sequences containing dynamic textures. Dynamic textures are video sequences with moving texture showing some stationarity properties over time, e.g. water surfaces, whirlwind, clouds, or crowds. By learning the temporal statistics of such content, corresponding areas in future frames can be synthesized. The encoder decides whether to use the synthesized content. The extension to this existing framework is twofold. First, we introduce another synthesis algorithm providing different content. Second, we present an additional macroblock mode which allows a very efficient signaling to the decoder to use the synthesized content. Simulation results show that a bitrate reduction of up to 30% over H.264/AVC at equal PSNR is achieved.