Local contourlet tetra pattern for image retrieval

Abstract

Local binary pattern (LBP) is an effective image descriptor that is being used in various computer vision applications such as detection of faces, object classification, target detection, image retrieval. To improve the performance of local patterns, different variants of LBP were introduced. In this work, contourlet tetra pattern, a modified version of local pattern, is introduced which uses contourlet directions to derive the tetra pattern of the image. The difference between local tetra pattern (LTrP) and the proposed method is that LTrP uses spatial first-order derivatives to derive the directions, whereas the proposed method uses contourlet transform to find the directions. In this work, contourlet transform is used to find the directions based on the fact that it helps to represent the images effectively into multiple directional bands which will have more accurate directional information than in the spatial derivatives. The proposed method is evaluated using three different databases (namely Corel 1 K, Corel 10 K and Brodatz), and experimental result shows the proposed method performs better than the conventional local pattern techniques.