Parallelizing Motion JPEG 2000 with CUDA

Abstract

Due to the rapid growth of Graphics Processing Unit (GPU) processing capability, using GPU as a coprocessor for assisting the CPU in computing massive data has become indispensable. Nvidia’s CUDA general-purpose graphical processing unit (GPGPU) architecture can greatly benefit single instruction multiple thread (SIMT) styled, computationally expensive programs. Video encoding, to an extent, is an excellent example of such an application which can see impressive performance gains from CUDA optimization. This paper details the experience of porting the motion JPEG 2000 reference encoder to the CUDA architecture. Each major structural/computational unit of JPEG 2000 is discussed in the CUDA framework and the results are provided wherever required. Our experimental results demonstrate that the CUDA based implementation works 20.7 times faster than the original implementation on the CPU.