Acceleration of a JPEG2000 coder by outsourcing arithmetically intensive computations to a GPU

Abstract

JPEG2000 is the new image compression standard for Digital Cinema. Digital Cinema is the upcoming technology for distribution of motion pictures for cinemas. The aim is to investigate the potential of accelerating JPEG2000 software coders in the context of Digital Cinema applications by outsourcing arithmetically intensive portions to a GPU (Graphics Processing Unit). The potential of accelerating JPEG2000 software coders with Nvidia’s GPGPU (General Purpose Computing on Graphics Processing Units) technology CUDA (Compute Unified Device Architecture) is investigated, taking restrictions stated in the Digital Cinema specifications into consideration. Implementation includes profiling and performance analysis of JPEG coders and integration into JPEG coder and eight types of Wavelet Kernels. /Kir09