Research On Image Mosaic Method Based On Measure Of Medium Truth Degree

Image mosaic refers to stitching some overlapping images of the same scene taken at differenttimes or in different locations into a complete image with a wide viewpoint. Under thebackground of MMTD (Measure of Medium Truth Degree), the key steps of image mosaic basedon feature points are further studied, in addition, based on the grey correlation model, theevaluation method of image mosaic is also studied by this thesis. The main content of this thesis isas follows:The traditional SIFT (Scale Invariant Feature Transform) algorithm has the problem ofinefficient and non-adaptability. To solve the problem, a SIFT algorithm based on MMTD isproposed. The algorithm adopts iterative method based on MMTD to determine the threshold.Experimental results show that the efficiency and adaptability is improved.The traditional RANSAC (Random Sample Consensus) algorithm has difficult to determinethe threshold. In view of this problem, a MMTD-based RANSAC algorithm is put forward. Thisalgorithm adopts MMTD to measure the distance between the pixel point which is obtained byaffine transformation and the corresponding matching point, then based on this distance, thethreshold is calculated in each process of iteration. Experimental results show that the newalgorithm has better matching effect compared with that of the traditional algorithm.A complete image mosaic based on MMTD is researched and implemented. FirstMMTD-based SIFT algorithm, NNDR (Nearest Neighbor with Distance Ratio) algorithm andMMTD-based RANSAC algorithm is adopted to obtain matching sets. Then MMTD is adopted tomeasure the similarity between matching sets. At last, a complete mosaic image is obtained byweighted fusion method. Experimental results show that the method obtains good mosaic effect.Grey correlation degree based on the fixed resolution has difficult to distinguish theimportance among indexes, and the evaluation results rely too much on the choice of resolution.Grey correlation degree of image mosaic evaluation method based on adaptive resolution isproposed. Experimental results show that the new method can distinguish importance of indexesaccording to the influence of each index to human visual system, which has a large differentiationfor mosaic effect of various algorithms and is reasonable and practical.