Image region extraction for content-based image retrieval

Dimitrios Androutsos, Konstantinos N. Plataniotis, Anastasios N. Venetsanopoulos

Abstract: We present a technique for coarsely extracting the regions of natural color images which contain directional detail, e.g., edges, texture, etc., and smooth color regions, which we then use for image database indexing. As a measure of color activity, we use a perceptually modified distance measure based on the sum-of-angles criterion. We then apply histogram thresholding techniques to separate the image into smooth color regions and busy regions where edge, texture and color activity exists. Color segmentation is performed on the smooth areas using HSV histogram techniques, to obtain image regions with one dominant average color. Database indices are then created from the busy regions using the directional detail histogram technique along with the color vectors representing the segmented smooth areas. Image retrieval is then performed using the histogram intersection method for the detailed areas and color vector distance measures are utilized for retrieval of the smooth color regions.