A color image retrieval method based on wavelet transformation G-Regions Of Interest

This paper in the research of Region-Based characteristic image retrieval method foundation, proposed a new kind of the color image retrieval method based on wavelet transformation G-Regions Of Interest (GROI). We first use HVS(Human Visual System) characteristic to choose the color space which fit for the visual characteristics, then use K-means clustering to extract the areas of interest in the wavelet transform domain, and using the local energy of the wavelet coefficients in the areas of interest as a texture feature, color´s mean and variance as a color feature, the barycentric coordinates as a position feature, we calculate similarity of between the image content and retrieval. The simulation results show that this GROI method which combined color characteristic, texture characteristic and position characteristic can more accurately find the necessary content of the images to users, significantly improve the accuracy in color image retrieval.