A Temporal Error Concealment Algorithm Based On Data Embedding

As the emerging of 3-G mobile communication system and booming of IP-based network, video communication wins great importance and prosperous future in multimedia services network today. Some successful applications have been very popular, such as multimedia email, interactive multimedia and e-commerce etc. At the same time, the error will come into being in the wireless channel or IP channel. Because of the limit in bandwidth of channel, data of video communication is always the one which has be compressed and coded, based on that, the data is so tender that the quality of picture will be worsen in error environment. In one word, reduction of raw video data rates is one of the key problems in wireless and grouping video communication.For effective video communication, there are two ways to avoid or recover the transmitting error. One is use of the error control and data recover in traditional data communication, such as FEC, ARQ and so on; its intention is to transmit data with no error. The other is that errors recover, such as signal reconstruction and error concealment, whose aim is to recover corrupted picture vividly.Because of the characters of human eyes, people can abide some distortion of picture or video. Based on that, corrupted picture can be recovered appropriately when video bit stream have the error, Toward the recover, it is not recovered accurately, but only conceals the picture, which will look like a whole picture. So it has the applicability in practice. For the reasons, error concealment has become to the main method in the domain of video communication. For decreasing the picture distortion owing to channel error, people attempt the whole bag of tricks. Nowadays, the technique based on encode, decode or combining of encode and decode has attracted people's attention.This paper describes the necessity and the challenges of error control for video communications, and provides an overview of various approaches that have been developed for in the current international video coding standard H.264. We emphasize the advanced coding features presented in H.264, such as multi-shape motion compensation block size, multiple reference frame prediction, fractional pixel motion compensation accuracy and context based entropy coding methods.To cope with the problem of packet loss probability and visual degradation, H.264 introduces slice-structured coding, which provides spatially-distinct resynchronization points within the video data for a single frame. A more advanced and generalized concept is provided by Flexible macroblocks ordering, which has been introduced to have the possibility to transmit macroblocks in non-scan order. This flexibility allows the definition of different patterns including slice interleaving without interrupting the inter macroblocks prediction for motion vector prediction and entropy concealment. A third error resilience concept included in H.264 is data partitioning, which can also reduce visual artifacts resulting from packet losses, especially if prioritization or unequal error protection is provided by the network. All these error resilience features in H.264 video coding standard can greatly provide more robust error resilient capabilities and keep compliance with original standard syntax.After a general analysis of the current situation of error resilience techniques in video coding, more words focus on the error concealments in the decoder. Generally speaking, the paper describes two aspects, one is the conventional error concealment, including spatial-domain, temporal-domain, frequency-domain; and the other is the up to date one. Nowadays, a lot of new technology turns up, this can be used in the video domain excellently, such as watermark. Based on that, people are thinking of error concealment making use of encoder and decoder.Based on the research above, this paper proposes an algorithm to enhance the robustness of constructed video in error prone channel. The algorithm is the error concealment based on data embedding, which, based on the new video standard-H.264, by the methods, the decrease between the same macroblocks of frames near by will be embedded in transform coefficients. Then in the decoder, when the program has searched the lost macroblocks, the embedded data is extracted from video, we use the information to find out the best block for concealing corrupted picture.The paper suggests two embedding methods, the one is embedded by the odd-even embedding; the other is embedded by fixed points embedding. The result shows that, compared with the algorithm in test model JM72, the two methods have the more cost in transmitting channel, and the PSNR also reduces in encoder. But compared with odd-even method with the least 0.10 dB, the other makes the less decrease: the least is only 0.01 dB, for the reasons, we ignore the PSNR reducing. Towards the PSNR in the decoder, two methods have the same results more or less, but the second method makes the less embedding cost. In one word, the method of embedded by fixed points embedding has the better effect.In conclusion, the proposed temporal error concealment algorithm has the effective improvement on error concealment. Besides, the performance is even superior using proposed algorithm as the QP or acuity-motion is increasing.