A search-aware JPEG-LS variation for compressed image retrieval

With increasing amounts of image data, such as the satellite images, being stored in compressed format, efficient retrieval of the images has become a major concern. For satellite images, one typical problem is, given an image pattern, we need quickly to locate its matches in the image. It is highly desirable that the compressed images are not decompressed when the matching is being performed. The problem is generally referred as the "two dimensional compressed pattern matching" problem. We report our recent work on compressed pattern matching in JPEG-LS compressed images. The method proposed is primarily a two-pass variation of the JPEG-LS compression algorithm and the algorithm is made search-aware. Experimental results show that the two-pass compression algorithm, unlike other search-aware compression algorithms, does not sacrifice compression. The proposed approach can be easily adapted for other compression methods such as lossless JPEG and CALIC.