Research On Repair Methods For Digital Video

Digital video repair is an important area research of video information processing, which can be used in video editing, transmission, storage, restoration and improvement of historical literature and materials, perfection and distinguish of video evidence and so on. From different points of view, this dissertation analyzes human visual perception and focuses on some key issues in the topic of video repair and it's application. The main work and innovations are listed as following:1. A 3D total variation-based video repair method is presented. First, an energy function for describing spatio-temporal smoothness for video is established. Then, ac-cording to variational method, the problem how to get the energy function extremum is converted to a problem how to solve an Euler-Lagrange equations. At last, combined with the digital video pixel distribution, a neighborhood weighted synthesis method is used for video repair. Because of the low complexity, the method is suitable for repair-ing small defect regions2. A method for repairing large region using motion information is presented. Affine model is used for describing the motion of non-repair region. Form the matching points extracts from neighboring frames using the scale-invariant features (SIFT) oper-ator, the global motion parameters is estimated. The region to be repaired is completed using the associated information extracted from the entire video by the motion param-eters. The method is applicable to repair the unknown region after object removal.3. A Navier-Stokes equations based video repair methods is given. Video can be viewed as three-dimensional spatio-temporal data stream. Based on the similarity on continuity and smoothness between the stream and the three-dimensional fluid, video data stream can be described with the Navier-Stokes equations which used for describ-ing the fluid continuity. From the continuity description, a local constraint equations is derived, which used for video repair. This method is suitable for small region, and can be used for renew the aged video literature.4. De-interlacing problem can be regarded as a video repair problem for small regions. In order to balance between accuracy and efficiency, several de-interlacing methods are given according to the distribution about known points and points to be repaired in the interlaced video. Experimental results show that these methods can take into account the space and time smoothness. The methods in this paper can mainly used for repairing video defects, and can also be used for repairing the video which has defects because of data lost in transmission and storage, and used for renew the aged video literature. Further research can include the distinguishing whether a video with or without repair, or which repair method has been used in a video material.