Study On Spatial Error Concealment Algorithm Based On Detail Optimization

With the implementation of 3G network, the importance of the online video application has been significantly increased in our daily life. During the transmission of the encoded bit stream of monoscopic or stereoscopic video over the Internet and wireless channels, packet loss occurs inevitably due to unreliable networks. Error concealment is an effective visual quality enhancement technique at the decoder by alleviating the degradation of the decoded images.The first frame of video sequence or group of images is I frame which is encoded by intra prediction, only the adjacent blocks around the lost block in the current frame can be used for error concealment. If the lost frame cannot be recovered well, it will result in error propagation to the following frames of the video sequence. For the monoscopic video sequence, I frame refers to the first frame of the sequence, which is a 2-dimensional image; for the stereoscopic video sequence, I frame refers to the first 3-dimensional stereoscopic image pairs, including the left view and the right view.In the typical video CODEC standards such as MPEG-4 and H.264, the error concealment for I frame of the monoscopic video sequence is the weighted average interpolation algorithm based on the median filter, using the adjacent pixels around the lost block to reconstruct the missing pixels. The method mentioned above can obtain good effects while operating on smooth areas of the image, but for the regions with a lot of details and textures will result in some image blurring. Recently many algorithms have taken into account edge detection and direction determination method and can recover some edge details of the lost block, but most of these technologies using theSobel operater, which is powerless for detecting linear edges with more details. Stereoscopic video technique is regarded as an important development direction of video technique in future. The spatial error concealment for the first stereoscopic image pairs has become a research focus. Most of the spatial error concealment technologies for stereoscopic video sequence is based on the disparity vectors generated from block-based stereo matching.With the improvement of living standards and the rapid development of high-definition display technology, people are not satisfied with the visual effect achieved by existing error concealment methods. Although the existing spatial error concealment technologies have had good results, spatial error concealment with more refined detail optimization is still a very challenging task.This paper has studied and implemented two spatial error concealment algorithms based on detail optimization. After summing up a large number of domestic and foreign literatures, analyzing plentiful simulation results, considering the complexity of algorithm and experimental results comprehensively, the spatial error concealment algorithm based on detail optimization has been put forward.In this paper, two new spatial error concealment algorithms based on detail optimization for monoscopic video sequence and stereoscopic video sequence are put forward and implemented in VC++6.0, respectively. The main task of this article is summarized as follows:(1) Spatial error concealment based on Hessian matrix with half pixel precision is proposed for monoscopic video sequences. In this case, associating with traditional spatial error concealment algorithm, we detect linear edges in frame by Hessian matrix, and implement half pixel interpolation to further refine the pixel information, obtain a much more accurate edge direction. During the direction extrapolation, instead of using the available pixels far from the to-be-concealed pixels, we use the half pixel points to improve the accuracy of the detection and optimize the reconstruction results.(2) Spatial error concealment algorithm based on detail optimization is proposed for stereoscopic video sequences. At decoding end, we implement stereoscopic macthing of damage channel and the other channel on a pixel-basis. Accurate disparity vectors can be obtained using the algorithm proposed in this paper. Firstly, the algorithm chooses a pixel as the matching unit, which can exactly match to a level of pixel. In this way, it provides a basic condition for better reconstructing the details. Secondly, the algorithm makes full use of the constraints of epipolar line, the more accurate stereo matching can be realized. Thirdly, the matching criterion we utilize the Normalized Cross-Correlation function, set a threshold to reduce the effections of occlusion and light intensity, and further improve the accuracy of stereo matching; set match windows and the search range to obtain much more accurate results.(3) The algorithm in this paper has achieved the purpose of having a better reconstruction result. From the subjective and objective evaluation of experimental results, the output view of H.264 decoder is fuzzy, especially in the detail zone, and also has some"block effect". The algorithm proposed in this paper can achieve better subjective quality and higher PSNR values.