A Comparative Analysis of Image Retrieval Techniques for Web Based Images

An image retrieval system is a computer system for browsing, searching and retrieving images from a large database of digital images. Web Based Image Retrieval addresses the matching and retrieval of images which share similar visual content to search a concept from a large database of un-annotated images. Matching images based on visual content is a fundamental problem in computer vision systems from a variety of fields including image retrieval and robot localization. Image retrieval is based on different types of features such as Edge based, Template based, and Dense Scale Invariant Feature Transformation (DSIFT). In this approach a new web based image retrieval system using DSIFT with Bayesian approach based segmentation technique are implemented. It performs three main tasks: extracting image features, comparing with database images and retrieving similar images from database. The Bayesian probability is calculated for DSIFT features to retrieve the most similar images. The image retrieval based on quantization with Bayesian based DSIFT retrieving system gives better accuracy compare to edge and template based image retrieval techniques. Keywords— content based image retrieval, edge based retrieval, Template based retrieval, Quantization with DSIFT, Bayesian classification with DSIFT.