Study On Order Self-adaptive And Detail Optimized Temporal Error Concealment Algorithm In H.264

With the development of network and wireless communication, thetransmission of real-time images through Internet and wireless channels has beenthe current research hotspot in the field of video communication. Various networkenvironments of the video streaming require more advanced techniques forcompression and transmission of images. H.264 is a newly developed videocoding standard which can be applied in broadcast and telecommunication andranges from low bitrate communication to HDTV. H.264 absorbed theexperiences from past video coding standards and the advanced coding technique.The coding efficiency of H.264 has been improved 2 to 3 times compared toprevious video standards such as H.263 and MPEG, due to its adoption of moreefficient technology and strategy. Meantime, H.264 can provide network friendlyinterface and error resilient methods in order to adapt to different transmissionnetwork. To some extent, H.264 is superior to all other video coding standards,and will be widely applied in the field of video coding transmission and Internetdigital television.Bit-error and data-loss may occur when compressed video data istransmitted on channel. Due to poor performance of channel, some areas of theimage can not be reconstructed because of the corrupted data. Therefore, the errorconcealment technique is needed for the decoder to reduce the influence to theimage quality by error.In the decoder, error concealment is not to correct the error codes andrecovery the original data, but to find out other relative data to replace the errordata by the correlation of the video image in temporal and spatial domain, and tryto conceal the lost data as exactly as possible so that the human eye can toleratethe distortion caused by error. It is emphasized that how to use the redundancy ofthe received video data in temporal and spatial domain when the error occurred,so that the decoding image can achieve acceptable quality. The process of errorconcealment is as follows: firstly, check the error codes by syntax and logic;secondly, search the resynchronization marker to attain the resynchronization;then, locate the error codes by locating methods in pixel domain;finally, estimatethe error data and conceal, which is the most important part.Error concealment in the decoder can be implemented in spatial, temporal orfrequency domain. In spatial and frequency domain, error concealment algorithmuse only the spatial correlation, that is, only refer to the neighbor correct blockand interpolate the error block. If there are more details in the images to beconcealed, it may be possible that the image is illegible and not correspondent tothe texture of the neighbor block due to the losing detail information. However,to the sequence image, the correlation between the neighbor frames in temporaldomain is very strong. So, motion estimation can be used to find out theappropriate image block to replace the error block and implement the temporalerror concealment. To the image sequence having strong correlation betweenframes, the visual effect of the reconstructed image using temporal errorconcealment is better than that of using spatial or frequency error concealment.Based on widely referring to the materials about video coding standards anderror concealment, the dissertation has discussed the present development in errorconcealment field and analyzed the major method in detail. The content of ourwork is organized as follows.Firstly, we combine the development of H.264 with the transmitting channelcondition of the compressed video data, describe error concealment technique ofvideo transmission, analyze the necessity of error concealment, and formulate ourresearch issues.Secondly, the commonly used error concealment techniques in the decoderare explored in depth. For the inter-predicted macroblock, the temporal errorconcealment is often used. In temporal error concealment algorithms, byestimating the motion vector, the error block is replaced by appropriate blocks inreference frame. Considering the complexity of the algorithm, only one referenceframe is used. The most simple method of motion vector estimation is to replacethe error macro block by the macro block which has the same locatio in itsprevious frame. This simple method is suitable for areas which have strongcorrelation among the video images. For areas with high motion, the effect is notso satisfied. Therefore, it is necessary to estimate the motion vector of the losingblock when reconstructing, which is based on the neighbor macro blocks (assumethat the neighbor macro blocks are not lost.) The concealment effect of themoving area can be improved by this method.Then, utilizing the features of H.264 adequately, the dissertation proposed anorder self-adaptive and detail optimized temporal error concealment algorithm . Itadequately utilizes the correlation between motion vectors, the1/4 pixel precisionin H.264 and the edge pixel information. Considering the losing macro blocksmay be on the edge of one frame, the algorithm chooses the concealment orderself-adaptively according to practical cases, so that there are more motion vectorswhich can be used for candidate of the lost macro block. This strategy is veryimportant, which can optimize the error concealment effect when the lostmacroblocks are at the edges.In addition, when estimating the motion vector of a lost macroblock, thealgorithm considers the blocks of 4×4 mode so that error concealment is moreprecise and the quality of the reconcstructed video can be improved.Finally, the dissertation implements the proposed temporal errorconcealment algorithm, and compare its performance with that using the defaultalgorithm in JM model. Simulation results show that the performance of theproposed temporal error concealment algorithm is equal or superior to that usingdefault algorithm. The maximum PSNR of reconstructed image increases morethan 1.5dB.