Collage error is a quantitative measure of the similarity between range block and "best-matching" domain block. It is relatively robust compared with the fractal encoding parameters which can be quite sensitive to changes in the domain block pool. However, up to now, fractal-based image indexing techniques are developed based on the fractal encoding parameters while collage error is overlooked. In the paper, we propose three composite statistical indices by combining histogram of fractal parameters with the histogram of collage errors to improve fractal codes based indexing technique. Experimental results on a database of 416 texture images show that the proposed indices not only reduce computational complexities, but also enhance the retrieval rate, compared to existing fractal-based retrieval methods.
展开▼
