2-dimensional motion estimation and its application in video frame interpolation

Digital video processing has become an important topic with widespread usage of multimedia applications due to the increasing processing power of computers. A video contains a considerable amount of data. Many methods thus have been developed to store and transmit videos efficiently. The ultimate goal of any video processing algorithm is to display the video with optimum visual quality. In some cases where bandwidth or storage capacity is limited, temporal sampling can be utilized to reduce the amount of video data. Video frame interpolation schemes have been developed to compensate for distorted visual quality due to the reduction in video frame rate by temporal sampling. Motion phenomenon which is a result of scene change in each video frame is exploited for video frame interpolation. In 2D motion estimation, it is assumed that objects in the scene undergo translational movement and new positions of these objects are determined approximately within the interpolated frame. It is required that motion vectors obtained from motion estimation algorithm for video frame interpolation should point to true object motion. In this work, the problem of calculation of accurate motion vectors has been dealt with. Known motion estimation methods are outperformed by the proposed method for the solution of the problem. Video frame interpolation has been chosen as an application of motion estimation. Keywords: Digital video processing, 2D motion estimation, motion compensation, block matching algorithms, video frame interpolation.