Image retrieval system based on sketches

In digital image processing domain CBIR (Content Based Image Retrieval) is an important research area. Many existing image retrieval tools such as Google Images and Yahoo! Image retrieve images based on the textual annotations given associated with images. With respect to these real time tools, the annotations are given with the help of keywords and text based search methods are used to retrieve the images later. The performance of such real world systems is not satisfactory. The purpose of a CBIR system is to take an image as input and retrieve most relevant images from database based on the color, texture and even shape of the given input image. This paper aims at discussing and developing a new CBIR system that not only works with input images but also sketches. Before processing such query pre-processing is done to fill the gap between the real world images and the hand drawn sketches. We studied concepts like SIFT, HOG and EHD. Experiments are done on two sample databases. The empirical results revealed that the proposed SBIR (Sketch Based Image Retrieval) is useful and can be used in real time applications. Index Terms – CBIR, SBIR, digital image processing, pre-processing