Analysis of Image Compression in Curvelet Domain

Abstract

Curvelet transformation is a multiscale representation of signals. It localizes signal in scale and position as wavelets, but it is the localization in orientation, which advances its performance in representing edges sparsely. Edges are the most crucial information in images, and efficient processing of the information containing data is the primary motivation for image compression processes. In this research work, Curvelet transform (CT) has been used in image compression. Even though, CT is a well-established mathematical tool, the literature suggests that no significant contribution has been in employing curvelets in image compression. In this work, the conventional compression standards like JPEG and JPEG2000 are studied extensively, and the procedure is extended and made compatible with curvelets. Experimental results show that the proposed method produces somewhat similar SSIM and PSNR values that of conventional standards, but improvement has noticed in the decrement of edge mismatch error but with poor performance in compression ratio. To mitigate this problem compressive sensing is used, which samples the signal in less number of samples and better sample reduction has been observed.