Video Frame Rate Up-conversion Based On Delaunay Triangle Mesh

The continuous development of ultra-high-definition TV and movie industry has driven people's demand for watching higher frame rate video,and the frame rate used in different countries and different devices is different,in order to achieve the frame rate between videos in these different standard formats and bring more shocking views to people,Frame Rate Up-conversion(FRUC)has been proposed and has become a research hotspot..Frame rate is the number of frames per second transmitted in frames per second(Frames Per Second,FPS).The higher the frame rate of the video,the smoother the motion in the picture and the finer the details,the better the viewing experience of people.The frame rate boosting technique refers to inserting some frames in the middle of the original video sequence,so that the original low frame rate video becomes a higher frame rate video.It is precisely because people have the need to obtain a better viewing experience,so FRUC technology has a very broad application prospects,it is worthwhile for scholars to spend time and energy in this regard to promote the development of this technology.Scholars from various countries have also proposed a variety of methods to improve the frame rate.The most widely used method is the FRUC technology based on motion compensation.This method fully considers the motion of moving objects in the video and makes corresponding motion compensation to obtain higher quality interpolated frame maps are therefore widely used in practice.This thesis first introduces the theoretical basis of frame rate upconversion technology and some common classical algorithms,and then focuses on a mesh-based method based on motion compensation FRUC technology.The main contents and achievements of this paper can be summarized as follows:1.Six classic image feature extraction and matching algorithms were selected for comparative analysis.It was found through experiments that the matching images of ORB+BF algorithm have more feature points and the highest correct rate.After RANSAC mismatch,it can reach nearly 100%.So this paper chooses this method for subsequent motion estimation and Delaunay Triangle(DT)mesh generation.Here,we propose a new three-step image matching algorithm: BRISK+FREAK+BF,the experimental results show that the algorithm has better scale invariance and illumination invariance.2.An improved DT mesh generation algorithm is proposed.The algorithm is based on the image matching experiment results of the above ORB+BF method.The feature points of the matching result are generated into a point set,and three pairs of feature points are combined for sixparameter affine transform,substitute it into the coordinates of the starting point to calculate the transformed coordinates to obtain the motion vector of the moving object;then generate the initial DT mesh,and insert the points in the point set continuously to determine whether the DT section is satisfied.The sub-criteria,if it is satisfied,delete the previous triangle and regenerate the new mesh.If it is not satisfied,continue to read the point set until the mesh generation ends.At the same time,the traditional empty circumscribed circle criterion algorithm is optimized.The traditional calculation of the center and radius of the circumcircle of the triangle,and then calculate the distance from the insertion point to the center of the circle to determine the positional relationship between the insertion point and the circle is converted into calculation.The cosine values of the two angles are judged by their magnitudes to determine the positional relationship between the insertion point and the circle.This greatly simplifies the complexity of the algorithm and greatly improves the efficiency.3.After generating the DT mesh,the image is motion compensated to obtain an interpolated frame map between the two frames.Here,the two frames we selected are the first frame and the third frame of the test video,so that the interpolated frame image can be compared with the original second frame image.This article is based on Windows7 + Visual Studio2015 + Opencv3.3.1 platform experiments,you can find,whether from the subjective visual effects,or objectively evaluate the Peak Signal-to-Noise Ratio(PSNR),Normalized Correlation(NC)and Structural Similarity Index Measurement(SSIM)of the interpolated frame image and the original frame image,can prove that the proposed frame rate improvement method is superior to the classical frame rate.The algorithm is improved and the implementation of the system is obtained.