A shape feature based image retrieval in DCT compressed-domain

In this paper, an approach of image retrieval directly based on shape feature in DCT compressed-domain is proposed. Firstly, according to the edge feature of DCT compressed image, the rough object shape is extracted from DCT compressed-domain to form a binary edge map; then the similarity measurement is redefined by using the moments invariant to improve the retrieval performance. It is supported by theoretical analysis and experimental result that this approach not only is fast, but also has a good performance. And it is also robust to rotation and translation.