| Digital TV is the trend of TV system. The post processing algorithms are the key development direction of digital TV. This paper mainly focuses on the three kinds of post processing algorithms in the digital TV. They are super resolution, detail enhancement and frame rate up conversion algorithms. And we try to do some profound research to simplify such three algorithms to make them that can be easily put into practice.The super resolution algorithm mainly focuses on UHD videos. We improve such algorithm in the aspect of facet phenomenon and fake high frequency in the flat area. We use principal component analysis to improve the facet phenomenon and adopt template of different directions with adaptive threshold. So as to easily hardware implemented, we simplify this algorithm to design several fast algorithms to meet the demands. It is with many practical values.The detail enhancement algorithm is mostly used in UHD digital TV. We propose a bold assumption, that detail layer of single image is the result of twice search and patch match under the framework of local self-similarity framework. However, over enhancement artefacts is improved by our method. It brings out fantastic outputs to nearly every pictures tested. We also change the search and match mode to satisfy the low complexity need in order to put into practical use.The frame rate up conversion is also related to UHD video. Based on the experiences above, local self-similarity model is combined with the traditional frame rate up conversion framework using 3DRS. It initializes the traditional motion vector fields in low resolution pictures. The refine and smooth operation are in the corresponding higher ones. We use the K-means operations in different sizes to delete the incorrect vectors. Only a quarter of data are used, but we can achieve the 95% of the original level. And the last but not least, it might provide a kind of solution way when hardware meets I/O bottleneck. |
PDF Full Text Request