A Research Of Optical Flow Algorithm Based On Block Matching

The instantaneous speed of the pixel projected by object in image is calledoptical flow which is the direct performance of velocity field. Through optical flow,we can get the motion and structure information of the object, and solve someproblems in pattern recognition subject, aviation and other field. Nowadays theresearch of optical flow algorithm is one of the hotspots in the field of visual motionanalysis.Optical flow based on block matching algorithm has some advantages such asfast operation speed, easy to implement, but also has some disadvantages such as lowaccuracy. In this paper, the block matching algorithm will be improved to make thedense optical flow image which is accordance with the motion object. In this paper,the research of the block matching algorithm is given below:1, In this paper, some solutions will be given for the distortion of block matchingcased by illumination change, the scale and cover object, respectively. We deduce ablock matching criterion based on gradient to solve the problem caused byillumination change,and put forward the reference image model changes over sizeand time to solve the matters cased by the scale and cover object.2, This paper will improve the block matching algorithm by combining withvariation theory. By joining the match into the energy functional, the algorithm canproved a dense optical flow image. This algorithm combine block matching andvariation optical flow algorithm, and prove dense optical flow image accordance withthe motion of the object.3, This article deduce the collection of the object and its projection in space andimage, and get the equation of the motion object in the image, and then get the motionobject detection method for obscured area combined with optical flow. This methodwarp the whole form of objects and can detect the obscured area of objects.