A comparative analysis of compression techniques - the sparse coding and BWT

The process of image compression has been the most researched area for decades. Image compression is a necessity for the transmission of images and the storage of images in an efficient manner. This is because image compression represents image having less correlated pixels, eliminates redundancy and also removes irrelevant pixels. The most commonly known techniques for image compression are JPEG and JPEG 2000. But these two have certain drawbacks and thus various other techniques have been popping up, of late. Recently, a growing interest has been marked for the use of basis selection algorithms for signal approximation and compression. In the recent past, the orthogonal and bi-orthogonal complete dictionaries (like the Discrete Cosine Transform (DCT) or wavelets) have been the dominant transform domain representations. But, the DCT and the wavelet transform techniques experience blocking and ringing artefacts and also these are not capable of capturing directional information. Hence, sparse coding method (by Orthogonal Matching Pursuit (OMP) algorithm) comes into picture. Another, novel technique that has taken up recent interests of the image compression area is the Burrows-Wheeler transform (BWT). BWT is generally applied prior to entropy encoding for a better regularity structure. The paper puts forth the comparison results of the methods of sparse approximation and BWT. The comparison analysis was done using the two techniques on various images, out of which one has been given in the paper. (C) 2016 The Authors. Published by Elsevier B.V.