SKETCH BASED IMAGE RETREIVAL SYSTEMS : A COMPARATIVE STUDY

Safiya P.C

Abstract: Image retrieval systems which uses image sketches as the input are known as sketch based image retrieval (SBIR) systems. Main advantage of SBIR system is that it does not need to have a high skill to draw a query sketch. One of the main challenges in sketch-based image retrieval (SBIR) is to compute the similarity between a query sketch and a database image. To deal with this problem, we propose a Sketch Based Image Retrieval approach by salient contour reinforcement. In the proposed method, the image contour is divided into two types namely global contour map (GCM) and the salient contour map (SCM). SCM is helpful to find out the object inside images which are similar to the query image. In addition to two types of contour maps we proposes a new feature called angular radial orientation partitioning (AROP) feature. AROP feature fully utilizes the edge pixel’s orientation information in GCM and SCM to identify the spatial relationships. Our AROP feature based on GCM and SCM is very effective to discover false matches of local features between query sketch and database images, and have a huge improvement in the retrieval performance. This paper presents a review and a comparison of few of the various SBIR methods commonly used in image processing. SBIR systems are very effective in the field of medical diagnosis, photos haring sites, digital library etc.

Keywords: SBIR, Salient Contour, Contour Reinforcement

DOI: https://doi.org/10.15623/ijret.2016.0531005