A study and optimization of the JPEG2000 image compression algorithm

Abstract

The aim of this project is to study and optimize JPEG2000 compression algorithm. JPEG2000 is a compression standard that uses wavelet transform. A color space transformation is first achieved, then the compression process in the wavelet domain is applied. Our purpose is to find the optimal color space, optimal wavelet, and optimal resolution level to compress any image. To carry out the experiments, the basic JPEG2000 algorithm is first implemented using MATLAB software. Then, tests are performed on different images selected from a free database named LIVE. For the study, three color spaces have been considered, Red, Green, Blue (RGB), Luminance/ Chrominance (YCbCr), and Hue/ Saturation/ Value (HSV). Furthermore, different wavelets and resolution levels have been considered. The compressed images are evaluated in terms of compression ratio, Detail Index (DI), and three quality metrics a full reference spatial metric named PSNR (Peak Signal to Noise Ratio), structural quality metric named MS-SSIM, and a blur metric (IQA).