Image Retrieval Method Based on Entropy and Fractal Coding

In content-based image retrieval system, describing and extracting image’s feature is a key question. An image can be characterized by its fractal codes, and fractal codes can be used as the image’s feature to retrieve the images effectively. This paper proposes a novel image retrieval method using information entropy and fractal coding. First, each image in the database is classified by computing information entropy which is compared with a given threshold estimated from the inquired image. Second, the inquired image’s fractal codes are generated via Jacquin method, which is applied to the same kind of database images with fractal tenth iteration decoding. Finally, the image retrieval result is obtained by matching the similar Euclidean distance between the inquired image and the iterated decoded image. Experimental results show that compared with the direct image pixels similar matching strategy, our scheme not only reduces retrieval complexity and retrieval time, but also guarantees the retrieval rate. Thus our proposed method is effective and feasible. Key-Words: Fractal coding; Image information entropy; Image retrieval; Similar