Enhanced Segmentation Technique for Image Retrieval
Muhammad Ali Riaz, Youngeun An, Jong-An Park · Proceedings of KIIT Summer Conference · 2009
In this paper, color features of the images are used to make a feature vector for image retrieval. Novel technique for segmenting HSV (Hue, Saturation and Value) is used in this proposed algorithm. We used different ranges of Hue and Saturation in HSV color space. Image is first classified and divided into A number of areas, and then each area is divided into S number of segments based on the number of color pixels it contains. The area which has more number of colored pixels is divided into more number of segments and the area which has less number of pixels is divided into less number of segments. This way we can construct the feature vector. Feature vector of the query image is compared with the feature vector of the database images and the results are obtained and sorted in ascending order. A web-based retrieval system is used for experimentation and results. Results show that even if we rotate the image or change the size of the images the proposed method gives robust and efficient results.