An enhanced CBIR using HSV quantization, discrete wavelet transform and edge histogram descriptor

Now a days, a great attention has received for content-based image retrieval by the researchers. It is very popular and interesting topic of computer vision. The basic requirement of content based image retrieval is to extract the appropriate information from the large image repository corresponding to query image on the basis of contents with better system performance. But, in the development of CBIR system with an appropriate fusion of low-level features is a big problem. This paper suggested an enhanced retrieval system by using the combination of HSV histogram, discrete wavelet transforms and edge histogram descriptor to extract the relevant information of an image. In this work, we used HSV color histogram with quantized non-uniform 72 bins to extract color information of image, discrete wavelet transform on each component (H, S and V) of HSV image to extract the complex texture pattern of image and global as well as local edge histogram descriptor on V component of HSV image to extract the geometry information of image. Here, Euclidean distance is used as similarity measurement to find out how similar the user image is with respect to the image in the database. For experimental analysis, 600 images of Wang image database are used and the results show that this approach gives good performance in term of precision and adaptableness while comparing with other's combining scheme.