Target Tracking Based On Local Non-similarity

Target tracking is a research hotspot in the field of computer vision,the purpose is to locate the target from the video sequence image without interruption.In recent years,target tracking technology has made great progress,is used in various fields,eg,military battlefield surveillance,traffic surveillance on civil,medical diagnosis et.Currently,tracking algorithm still has to be solved:1.The tracking algorithm is high in complexity and time-consuming,and it is impossible to track in real time;2.The target undergoes deformation,scale change,and occlusion during the movement,and the target is lost,the tracking is invalid.So it is difficult to find a robust,general-purpose algorithm.In this paper,the paper studies and analyzes the above problems,and proposes a simple and applicable algorithm-target tracking algorithm based on local non-similarity.This article mainly starts from the following aspects:1.A new feature selection strat-egy based on local non-similarity is proposed,this strategy is stable and accurate when tracking.2.When performing block matching,use integral image to reduce the complex-ity of the algorithm,use parallel to improve the operating efficiency of the algorithm.3.When sorting,proposed a new previous M item sorting method,can greatly improve the efficiency of the previous M item sorting.4.Apply our calculation block matching algorithm to BM3D noise reduction,improve the efficiency of BM3D noise reduction.