The Robust Analysis Of Variational Optical Flow Algorithm And Its Application In Medical Image Processing

Optical flow is a research focus of computer vision and image processing, sincethe optical flow contains not only the movement parameters of the object or scene,but also the abundant information of structure. Variational optical flow algorithm hasunified model and high precision optical flow estimation. It is the most developed,and also the most complete algorithm among several types of the mainstream opticalflow algorithms. While the robustness analysis of optical flow algorithms is oftenassumed, or verified empirically, lacking of theoretical support. In this paper, therobustness of variational optical flow algorithm is analyzed theoretically and it isapplied to medical image processing.In the beginning of this paper, the theoretical knowledge of the variationaloptical flow algorithm is introduced, and the detail classification of the commonconstancy assumptions of data terms and regularization methods were discussed.Meanwhile, the advantages and disadvantages of the method and the applicability ofthe data terms and smooth terms were introduced. After all, three commonvariational optical flow algorithms were briefly analyzed.By analyzing the whole process of algorithm implementation, an universalrobustness analysis method of variational optical flow algorithm is proposed.Studying a robust variational optical flow model, convex energy functionals makesure a unique minimum, and this global minimum can be found in a stable way byusing standard techniques from convex optimization. In the linearization anddiscretization process of the energy functional corresponding Euler-Lagrangeequation, carse-to-fine technique is combined to avoid getting trapped in a localminimum. The SOR iteration method is proved to be convergent by algebraicproperties of the matrix expression of the system of linear equations. According tothe properties of special matrix, the error bounds of SOR iteration method isextracted. Under certain conditions, the SOR iteration method is numerical stabilityand the system of linear equations is well-behaved. Iteration expressions of commondata items and five smooth items corresponding diffusion equations shows that therobustness analysis method has a certain universality which proposed in this paper. The optical flow of different five groups image sequences is calculated inexperiment, and experimental data prove that robust variational optical flowalgorithm able to handle well multiple types of image sequences. The white gaussiannoise is adding into input images.According to the analysis of experimental results,the conclusion can be obtained easily that robust variational optical flow algorithmhas better antinoise ability and robustness.The high precision variational optical flow algorithm is used to calculate theoptical flow field in four groups image sequences, including MR image sequence,CT image sequence, video image sequence and microscopic image sequence. It isverified that variational optical flow algorithm has the ability to analysis the differenttypes of medical image sequences dynamically.