A combination of image retrieval and compression

An embedded content-based image retrieval algorithm is proposed in this paper, which involves extracting texture features from JPEG-LS, the latest JPEG lossless image compression standard. Considering the fact that prediction in JPEG-LS is conducted in a template of four surrounding pixels and local texture is analysed during the process of detecting edges before the predictive value is produced, a number of states specifying the feature of local texture is extracted to form a histogram as one of the keys to index each image at the population stage. Apart from the advantage that compression is embedded inside the indexing and retrieval, experiments on a database of 2715 images with variety of content and sizes also show that, compared with the existing texture-based techniques, the proposed provides a better performance in terms of psyco-physical test on those retrieved images.