Visual content based clustering of near duplicate web search images

Near-duplicate detection has received substantial attention over the past few years due to applications in copyright enforcement, organizing large image databases, increasing focus in image search, duplication elimination of logos, saving storage space by removing redundancy, etc. In case of document images, near-duplicate detection can be used to increase the efficiency of tagging the documents by reducing the need for manual inspection of the documents. In this paper, an approach is presented to detect near-duplicate images using feature extraction and clustering process. Initially as a preprocessing step, noise removal and image enhancement is done. Image features are used for feature extraction and also for clustering the images. Appropriate similarity measure is used in accordance to the clustering algorithm. Clustering of images is performed which is followed by its evaluation. From the result of evaluation, the clustering process is refined to get better clusters. Each of these clusters will have one image as a representative of that cluster and other images in the cluster is called its near-duplicates. Finally performance measure is calculated for evaluating the algorithm accuracy.