Correlation and Probabilistic Relaxation Image Matching Approch

The statistics of the neighborhood gradients in an image can yield useful descriptions of target features. This paper presents a new approach on auto-correlation relaxation algorithm by using those features to match image clips. This method is iterative and begins with the detection of all potential correspondence pairs. Each pair of matching points is assigned a number representing the probability of its correspondence. The probabilities are iteratively recomputed to get global optimum sets of pairwise relations. This method could be found wide applications of matching video frames, and industrial detection recognition, for example in the research of artificial life and factory automation.