ANALYSIS OF IMAGE RETRIEVAL TECHNIQUES

Abstract

In the recent years we have observed a rapid rise in the size of digital image compilations. Giiga bytes of test images are created daily by both civilian as well as military equipment. However to manage such incredible amount of data pouring in everyday from so many sources we need to have an efficient storage and retrieval system for images. Since the 1970s work on retrieval of images is being actively carried on. Image retrieval can be looked upon from two angles- one his text based and the other being visual based. We have divided our work in two phases – first we have tried to retrieve text from an image using Support Vector Machines (SVM) and Maximally Stable Extremal Regions(MSER), second we retrieve image using the Bag of Visual Words technique (BoVW) where we have worked on inscription images. The text retrieval is carried on by first getting the output of images containing text from the SVM and then further processing of the SVM result using MSER. The proposed method for inscription image retrieval can be used to recognize inscriptions in languages from across the world. SURF (speeded up robust features) is used as an image feature extractor. A visual vocabulary is created by representing the image as a histogram of visual words which helps in the retrieval process. Usage of SURF ensures scalability, faster processing better results with darkened and blurred images. We demonstrate the method on a combination 300 inscriptions images comprising of several languages.