A Parallel Unit Encoding Stage for BCWT using GPGPU

E Sudarshan; Ch. Satyanarayana; C. Shoba Bindu

doi:10.37591/joipprp.v4i3.1125

Abstract

AbstractA backward coding of wavelet trees (BCWT) algorithm begins the coding process from the lower-level to the higher-level of the wavelet tree. While processing, it builds up the maximum quantization levels of the descendant (MQD) map and maximum quantization level (MQL) by the support of the wavelet coefficients of the corresponding image. In the processes of making the serial encoding stages into parallel stages, primarily we deal with a prominent prospect called parallel unit encoding. The parallel unit encoding stage will save the memory up to 30% by avoiding the contiguous zero bits from the original element bit-string before storing into the output stream. The algorithm velocity improved up to 17% by the inclusion of the parallelism on the development of the GPGPU platform using CUDA architecture over the CPU based BCWT encoding procedure. Keywords: Parallelism, BCWT, MQD map, CUDA, maximum quantization level, GPGPUCite this ArticleSudarshan E, Ch. Satyanarayana, Shoba Bindu C. A Parallel Unit Encoding Stage for BCWT using GPGPU. Journal of Image Processing & Pattern Recognition Progress. 2017; 4(3): 13–23p.