Image Retrieval System Using Kirsch based Local Ternary Pattern

Abstract

This paper addresses the challenge imposed \nby the tremendous growth of digital data to retrieve \nrelevant images. In this paper, a novel feature design \nmethodology is proposed to represent images efficiently \nfor Content Based Image Retrieval (CBIR). Gener- \nally, local patterns are computed directly on images \nand hence, directional information of the images is \nignored. In the proposed feature, Kirsch operators \nare used to highlight eight major directional changes \nin the image and further, Kirsch Ternary Local Pat- \ntern (KLTP) is extracted by analysing local inten- \nsity variations in the neighbourhood. In KLTP, global \ncolor information is also incorporated to make it robust \nand perform well on variety of images. Experiments \non natural and texture databases are done to verify \nthe performance, as compared to the available features \nin the literature.