Compressed Pattern Matching for Predictive Lossless Image Encoding

Pattern matching in compressed image domain is a new topic in computer science. Many works have been reported for pattern matching for compressed text and for lossy compressed image. However, searching of images in lossless compressed domain is almost a blank area and needs to be explored. Lossless image compression is widely used in areas such as medical images, satellite images, geometric images and many other areas that need to losslessly maintain the data of the images. Being able to searching in the compressed domain will save disk space and searching time and bring up considerable economic savings in these areas. In our work, we have studied the possibility of compressed pattern matching for the most three popular lossless image compression schemes: lossless JPEG, CALIC and JPEG-LS. Our study indicates that these algorithms can be search-aware by minor modification. We also present a modified JPEG-LS algorithm and the corresponding searching algorithm. Experimental results show that our method, comparing with the “decompress-then-searching” method, has nearly 30% improvement in searching time for most natural images. The modified JPEG-LS algorithm also has shorter encoding and decoding time, with an improvement of about 12-15% and 812%, respectively, for most natural images. The tradeoff is the decrease of compression of about 2% -8%. To our best knowledge, this is the first report on JPEG-LS compressed matching algorithm and this is the first “competitive” compressed pattern matching algorithm for lossless image compression.