Efficient object tracking using local tetra pattern based texture feature

Abstract

This paper addresses an efficient mean shift object tracking algorithm, by employing the joint color-texture histogram for object representation, and the mean shift algorithm for object tracking. The textural information of the object is extracted by using the four directional code called the local tetra pattern. The Ohta color model has been used for extracting color information. The local tetra pattern has the ability of extracting more detailed information as compared to other patterns such as Local Binary Pattern (LBP) or Local Ternary Pattern (LTP). The combination of the local tetra pattern-based texture information and Ohta color information makes the object tracking technique robust and efficient. The performance evaluation of the proposed algorithm has been carried out for different exigent conditions such as abrupt motion of object, varying illumination, clutter and occlusions. The qualitative and quantitative performance comparison of the same has been carried out with other existing algorithms by considering the detection rate, tracking speed, average precision, average recall and F-measure. The simulation study reveals that, the proposed method overthrowing the challenges like partial occlusion, non-rigidity and illumination variation.