視覺(jué)穩(wěn)健跟蹤技術(shù)及應(yīng)用研究.pdf

1、The Structural SIMilarity Measure (SSIM) combined with thesequential Monte Carlo approach has been shown to achievemore reliable video object tracking performance, compared withsimilar methods based on colour and edge hi

2、stograms andBhattacharyya distance.However, the combined use of the structuralsimilarity and a particle filter results in increased computationalcomplexity of the algorithm.In this paper, a novel fast approachfor video t

3、racking based on the structural similarity measure ispresented.The tracking algorithm proposed determines the state of thetarget (location.size) based on the gradient ascent procedure appliedto the structural similarity

4、surface of the video frame, thus avoidingcomputationally expensive sampling of the state space.The newmethod, while being computationally less expensive, has shownhigher accuracy compared with the standard mean shift alg