Research Of Error Concealment For H.264 Video Stream

With the development of network and communication technology, the video communication is applied more and more widely, including visual telephone, distance education, video on demand business, video monitoring etc. In order to reduce the requirements of the channel bandwidth, image, video or other multimedia must be take an efficient compression coding. Due to the decrease of temporal redundancy and spatial redundancy of video image sequence, it is easy to cause the compressed bit stream more sensitive to the errors of channel. It may be lead to error diffusion of the time domain and spatial domain to any bit errors or missing so that the quality of the received video images suffering from the serious influence. Therefore, error concealment techniques for H.264 compressed video stream has important signification to the transmission in the channel.The paper is committed to research on the spatial error concealment project for H.264 video stream. The proposed method outperforms the standard algorithm based on subjective and objective measures. Study in detail is as follows:(1) Aiming at the H.264 standard algorithm questions that the smoothness of restored video is poor to the complex-texture block, the project of adaptive error concealment based on multi-texture direction is proposed. Using two thresholding set the thresholding of Sobel algorithm, combined with cost function of the boundary pixel difference to determine the interpolation direction for each pixel of error block. Judging the thresholding of Sobel edge detection to detection the edges of adjacent, then, recovering the error block by the interpolation direction of each pixel. The proposed method improves the smoothness of restored video.(2) H.264 compressed video stream occurred easily to data loss and bit errors in the transmission process, at the same time, face region in decoding reconstruction was fuzzy. Aiming at these questions, this paper proposed an error concealment algorithm based on adaptive directional interpolation and facial region of similarity constrains so as to carry on error concealment two times to error block. First, used spatial error concealment algorithm based on adaptive directional interpolation to recovering the error block by an interpolation to achieve the first error concealment. Then, it extracted the foreground object of error block to gain the target face region, using the facial partition method to locate the possible face region and calculating the similarity between adjacent frames and blocks. Last, after the affine transformation, corrupt the face region of error block used gradient information of the target face region to achieve the second error concealment.In summary, the paper respectively proposes an improved project between the fields in video image reconstruction of texture information complex and the face region fuzzy after decoding. At the same time, JM8.6 simulation experiments show that the proposed method outperforms the existing algorithm. Therefore, it has more extensive applicability.