Flexible color texture retrieval method using multi-resolution mosaic for image classification

A color texture retrieval system has been developed using a multi-resolution mosaic for flexible image classification. First, a representation of the color texture image is investigated. The texture can be characterized by features such as shape, structure, color and randomness. The texture features are extracted using operators. Then the feature images are transformed to lower dimension feature vectors using multi-resolution mosaic processing. Next, a similarity function is calculated for an unknown input color texture image. Finally, a new distance, with weight functions, is calculated using the similarity functions, and the results are sorted. By selecting weight functions, we can reflect the impression of texture features flexibly in retrieval. The effectiveness of retrieval is demonstrated in simulations with a database of several color texture images.