DSP Implementation of Wavelet Transform Based Embedded Block Coder Using Quadtree Partitioning Method

Abstract

AbstractThis paper work describes the implementation of embedded block coder for still image compression using only quad-tree partitioning method. This technique is based on Discrete Wavelet Transform. The motivation behind this work is from SPECK (Set Partitioning Embedded bloCK) algorithm. It uses a recursive set-partitioning procedure to sort subsets of wavelet coefficients by maximum magnitude with respect to thresholds that are integer powers of two. The proposed method simplifies the complexity of the embedded wavelet image coding algorithm by combining both sorting pass and refinement pass. In comparison with other methods, this is simpler to be realized on hardware and has higher compression efficiency. This paper work also explains the software and hardware implementation by using TMS320C6713 DSK board. The statistical analysis is done with profile statistic available in Code Composer Studio (CCS) environment. The MATLAB simulation results show that PSNR values are quite improved by lowering threshold values.KeywordsImage compressionDiscrete Wavelet Transform (DWT)Quadtree PartitioningTMS320C6713 DSK BoardPSNRMSE