Research On Optical Flow Calculation Technology Of Image Sequence Based On Guided Filter

Optical flow is refers to the objects in 2d or 3d scene projection plane pixel motion of the instantaneous velocity,it’s not only carried by watching the movement of the object information,and also contains the object’s 3d structure information.So image sequence optical flow computing technology research is image processing,pattern recognition and an important research direction in the field of computer vision,etc.Since the 21 st century,with the rapid development of computer hardware and software and display technology,image sequence optical flow computing technology has been in various fields of social production and life highlights the increasingly important status and role,research results are widely used in weather analysis,traffic video monitoring,unmanned automatic driving system,medical image segmentation and robot vision system,etc.In recent years,with the deepening of optical flow calculation,the new theory and new technology for optical flow calculation model and optimization method have been emerging,and the image sequence optical flow calculation technology has made breakthrough progress in calculation accuracy and calculation efficiency.However,the problem of accuracy and robustness of optical flow calculation needs to be further studied when the image sequence contains difficult scenes and motion forms such as blurred motion,non-rigid motion,brightness mutation and discontinuity.In this paper,we focus on the problem of motion blur and so on.By designing the optimization model of image flow based on guided filtering,this paper overcomes the optical flow calculation caused by complex edge structure,large displacement and non-rigid motion in the current image edge blur and other issues,in order to improve the accuracy and robustness of optical flow calculation.The main work of this paper includes the following points:1.The research background and research status of the image sequence optical flow calculation technology are introduced in detail,and the main problems and challenges of the optical flow calculation technology are analyzed in detail.2.The typical optical flow calculation methods such as HS and CLG-TV are introduced,and the representative optical flow calculation models of Classic+NL and NNF are introduced in recent years,and the advantages and disadvantages of each method are analyzed.3.Aiming at the problem of complex edge structure,large displacement motion and edge blur caused by non-rigid motion,an optimization method of optical flow based on guided filtering based on hierarchical refinement of image pyramid is proposed.In the process of optical flow calculation,the details of the edge of the image can be preserved by the linearization of the guided filtering,which can overcome the problem of image edge blur and make the edge of the image reach the effect of enhancement and sharpening,and improve the optical flow computational accuracy and robustness.4.The standard test image sets provided by Middlebury,MPI Sintel and KITTI databases are used to optimize the HS,CLG-TV,Classic+NL and NNF methods.The experimental results show that the optimization model of optical flow based on guided filtering can effectively improve the accuracy and robustness of optical flow calculation,especially when the image sequence contains complex edge structure and large displacement movement,can greatly enhance the effect of optical flow calculation.