Performance Evaluation of Texture based Image Retrieval

Abstract

This paper presents a novel content based image retrieval (CBIR) system based on Haar Wavelet Transform. Content Based Image Retrieval (CBIR) has been an active research area. The CBIR is to retrieve the images based on a query image, which is specified by content, from the given collection of images. Current system uses texture as a visual content for feature extraction. The present work uses modified Haar wavelet transformation for feature extraction of an image. Here Haar wavelets constructs the feature vector of size ten, characterizing texture feature of the images, only in three iterations of the wavelet transforms. The K Means Clustering Algorithm is then used to cluster the group of images based on feature vector of images by considering the minimum Euclidean distance. The performance evaluation of the present method is done by Precision and Recall for different databases.