The Research And Implementation Of Video Inpainting Technology

Research on the theory and algorithms of video inpainting have attracted remarkable attention in recent years.Video inpainting is a kind of technical which is implemented on the plausible effects to remove the interesting objects from the video,or to fill missing parts into a video sequence.The missing parts mentioned here is general:it may be still or moving,in the background or in the foreground,it may occlude one object or may be occluded by some other objects.At first,the definition of video inpainting is introduced briefly,then,three important video inpainting methods are discussed and improved,seperatedly,as follows:The improved SPREF based video inpainting method use an improved oct-tree model to segment video,improving the precision of segmentation during inpainting.By adding the SPREF components into SSD,the improved video inpainting algorithm based on temporal correlation is free of the objects whose scale have been changed during the inpainting.The proposed methods changs the inpainting sequence of corrupted frame, following the sequence to inpaint from back and front to middle simultaneously.The improved video inpainting method based on video epitome use multiple scale principle to learn the video epitome,allowing us to work with multiple size patches in our experiments,from very small patches capturing fine detail to very large patches that help strengthen the long-range correlation.