A fuzzy quantization approach to image retrieval based on color and texture

摘要

This paper presents a new image retrieval method which is based on color and texture features. By using fuzzy quantization (which is based on a linear subjection function in the quantization of HSV color space), this method attempts to make the quantization results more accessible to human perception; furthermore, according to the information of the extracted dominant color of partition, we introduce a neighborhood color matrix which is used to describe the relative color spatial distribution, for the purpose of improving the robustness of image transfiguration. With the supplementary information of image textures, our method combines both the image and texture features to conduct composite image retrieval. Our experimental results show that this method can greatly improve the retrieval accuracy.