A Comparation Between SVD and SSVD Method on Image Compression

Abstract

Data compression is the process of converting an input data into other data which has a smaller size. The data can be a file on a computer or a buffer in the computer’s memory. Compression is useful because it helps reduce the consumption of resources such as data space or transmission capacity. The use of singular value decomposition (SVD) in image compression has been widely studied. If the image, when considered as a matrix, has low rank, or can be approximated sufficiently well by a matrix of low rank, then SVD can be used to find this approximation. That is, by not include some elements of the image, the approximation has been able to represents the original image. This thesis presents a variation for SVD image compression technique proposed by Ranade et al. called SSVD. This variation can be viewed as a preprocessing step in which the input image is permuted by independent data permutation after it is fed to the standard SVD algorithm. Likewise, this decompression algorithm can be viewed as the standard SVD algorithm followed by a postprocessing step which applies the inverse permutation. On experimenting with some standard images, SSVD performs substantially better than SVD. This thesis also presents experiment evidence with other simulated images, which appears that SSVD isn’t better than SVD.