Shape Features Extraction Method for Content based Image Retrieval

Abstract

Shape identification is very important and one of key operations in context of image matching, search and retrieval. If shape is recognized properly, there is no more difficulty remained while comparing two images and deciding they are either similar or dissimilar. This research is completely dedicated to build up a new shape feature extraction method that is employed in content based image retrieval (CBIR), for efficient and accurate image search and retrieval. Firstly, image segmentation has been performed for isolating the shapes appearing in images. Secondly, six standard shapes circle, ellipse, hexagon, triangle, square and rectangle have been brought into use and their tangent coefficients have been calculated. Thirdly, the sum of squared difference has been calculated to obtain the image matching key. Finally, color difference has been calculated to measure the color attributes of two images. Comparison has been performed with some of existing latest methods in terms of average precision to show the better and improved performance.