| With the improvement of information technology and the increasing development of society, the more actual applications such as wireless multimedia communication, multimedia email, interactive mutimedia, remote medical and e-commerce has become the next focus of people. From the viewpoint of technology, video compression with low bitrate is the key point to implement these applications. It puts forward the higher challenge to compression efficiency, error control and error concealment, object coding and content-based interactivity. The topic of this thesis is proposed under this premise.The current transmission channel is with noise, thus it will appear the scenarios of lossing data packets. Therefore, the error free transmission in the actual network can not be guaranteed. It is inevitable to encounter error or loss of bits in the encoded bit stream. In order to ensure the service quality of video transmission, error control is needed during video transmission to reduce the affection of error. One has put up various efforts in order to reduce distortion caused by channel error. Thereinto, besides the well known method such as FEC and ARQ, various error concealment methods based on decoder or decoder combined with encoder have been noticed by researchers.Error concealment is to detect error and skip the part of bitstream which can not be decoded due to error and proceed the decoding process when error appears in the vdieo bitstream and recover the image quality of the regions which can not be decoded to the greatest extent. This method is mainly suitable to the video images of real time transmission. It is a type of make-up method for the occurred error, thus it is also a more independent and simply implemented method, which has much effect on improving viewing quality of error images. This technology uses certain methods to weaken the impact of error on human view when detecting errors. It does not really eliminate the error in transmission. The technology of error concealment recovers the lost or damaged data at the decoder by making used of the obtained correct image information and their inherent temporal or spatial correlation.Error concealment techniques can be classified as the following two categories: the spatial domain error concealment and the temporal domain error concealment.The temporal domain concealment is based on the temporal relativity,this technique is adapted to the video signal that motions not very complicatedly.However,in the case of scene switch or background exposure,it is difficult to find a suitable block to replace the damaged one from the forward frame.Therefor,we need to deal with the damaged block using intraframe information.The spatial error concealment can not only carry out the intraframe concealment,but also reconstruct some boundary information.The purpose of our work is to investigate a spatial error concealment which detects edges based on Hessian matrix and use directional interpolation.The production of this investigation can be applied in multimedia communication, video telephone,image and video frequency searches,and can receive more acceptable vision quality via error concealment.Furthermore,participating in some other international or national standard code techques,this technique can form a code technique which belongs our country's own intellectual property rights.Hence,this task have its positive realistic meaning.In this paper ,we propose a spatial error concealment algorism which based on sub-pixel edge detection with Hessian matrix,and classifies the to-be-concealed macroblock by the pixels in it as follows:the pixels in macroblocks which have many linear edges,the pixels in macroblocks which have many grads edges and the pixels in macroblocks which is smoothness.For the one which have many linear edges,we adopt the sub-pixel edge detection with Hessian matrix to heighten the reconstructed effect of the details of the image;For the one which have many grads edges,we adopt edge detection with Sobel operator to reconstruct the important edge information;and we adopt weighted pixel average for the one which is smoothness.This method employs a spatial interpolation with directional extrapolation.The ascertainableness of the direction is differ from the former edge detection on the pixel level.This paper will put forward a new sub-pixel level edge detection.The precision of edge detection effects directely the exactness of the directional extrapolation,as a result,it is necessary to apply the edge detection on sub-pixel level for the images which have more linear details.The mostly merit of this edge detection on sub-pixel is that it heightens the precision.Therewith,at the same time,along with the enhance of the precision of the detection,we cannot but considering a problem which is the increase of the complexity of the operation. Essentially, Hessian edge detection technique proposed in this paper still belongs to one kind of stimulant interpolation algorithms,but it reduces the complexity of the algorithm due to the use of Hessian matrix.Therefore,this method processes preferable performance and be better used in common currency.After detecting the direction,we apply a method with point-to-point extrapolation to implement error concealment,and employs a consecutive reconstructed mode to conceal the losed blocks.According to this method,those pixels which have reconstructed already can be applied to the to-be-reconstructed pixels in the follows easily.The mainly merit of the consecutive reconstructed mode is that the reduce of the statistical model's complexity,namely replaced the block-based to the pixel-based,and hence enhanced the capacity of the reconstruction of the important image characters.In addition,aimed at the error concealmeng inter macroblock,different conceal sequence will result in different outcome.This paper proposed a sequence of from outer to inner,layer upon layer,and have validated that its impression is quite ideal.In this paper,we put up a mass of experiments to the spatial domain error concealment based on sub-pixel level edge detection with Hessian matrix,and validates the algorithm in our paper via comparing with former error concealment methods. |
PDF Full Text Request