Research On Algorithm Of Low Complexity Frame Rate Up-conversion For Ultra-high Resolution Videos

Frame rate up-conversion(FRUC)is also called temporal interpolation of video,which is a video post-processing technique and can improve the subjective quality effectively.At present,this technology has a very wide range of applications:it can improve the viewing quality by increasing the refresh rate so as to reduce trailing and motion blur at the edge of object in liquid crystal display(LCD)when displaying moving scenes;FRUC technology can also solve frame rate conversion under different video format standards,so as to watch videos among various multimedia terminal equipment;in the low bit rate video transmission system,this technique is required to recover the droped original frames as much as possible in order to restore the original video.This paper will study the block-based motion compensation FRUC algorithm.Under the framework of bidirectional motion compensation,key designs are put forward for the bilateral motion estimation,including hierarchical strategy,candidate vectors selection,jitter of the motion vectors and the processing of boundary blocks.It can not only reduce the computational complexity,but also improve the accuracy for big motion vector and boundary motion vector.Meanwhile,motion vector field global scanning and adaptive motion vector smoothing were put forward to improve the convergence in moving boundary,and enhance the accuracy,consistency and denseness for the motion vector field.In the result,the quality of the interplated frames were improved with low complexity implementation.Based on the bilateral motion compensation,an occlusion processing strategy was proposed aimed at improving the quality of FRUC.Through the analysis of occlusion,we propose to detect the occlusion areas by pixel difference,divide the types by moving difference,and implement one-direction motion compensation based on the occlusion type for cover or uncover regions,therefore,the false contour of the object boundary can be effectively alleviated and better quality of the interpolated frames can be achieved.Some problems met in the applications of FRUC are analyzed and the corresponding solutions are proposed.For the scene conversion problem in the video,a scene change detection scheme based on the mitching error results got in motion estimation stage is presented in this paper,so as to generate the interpolated frame by non-motion compenstaed algorithm and a smooth transition can be achieved.For the scrolling subtitle problem,a full analysis of the vector field was carried out and the global motion vector in the text regions would be estimated combined with the specific properties of subtitles.As a result,the interpolation result are improved.