GPU-friendly EBCOT variant with single-pass scan order and raw bit plane coding

摘要

A major drawback of JPEG 2000 is the computational complexity of its entropy coder named Embedded Block Coder with Optimized Truncation (EBCOT). This paper proposes two alterations to the original algorithm that seek to improve the trade-off between compression efficiency and throughput. Firstly, magnitude bits within a bit plane are not prioritized by their significance anymore, but instead coded in a single pass instead of three, reducing the amount of expensive memory accesses at the cost of fewer truncation points. Secondly, low bit planes can entirely bypass the arithmetic coder and thus do not require any context-modelling. Both the encoder and decoder can process such bit planes in a sample-parallel fashion. Experiments show average speed-ups of 1.6x (1.8x) for the encoder and 1.5x (1.9x) for the decoder, when beginning raw-coding after the fourth (third) bit plane, while the data rate increases only by 1.15x (1.3x).