| Video transcoding is a core technology for providing universal multimedia access by converting one compressed video to another.It is widely used in video conference,live online and remote education.However,in many practical transcoding applications,the source videos are often contaminated by noise during acquisition,storage and transmission.So it is necessary to add denoising algorithm to video transcoding.The noise video transcoding is a technique for converting the input noise video to denoised video by adding a denoising operation.Providing improved video quality for video analysis and processing.Noise video transcoding has an urgent need in video surveillance and mobile video communications of the application requirements.The basic idea of fast video transcoding is to improve the transcoding speed by using the correlations before and after transcoding,such as macroblock mode,motion vectors and so on.But in noise video transcoding,the correlation between noise before and after transcoding becomes complicated.The traditional fast transcoding algorithm does not take into account the impact of noise,resulting in noise video transcoding mode mapping error,and the final transcoding effect is not good.To solve the problem,the paper analyzes the influence of noise in inra mode and inter mode,and establishes a pattern mapping rule for noise video transcoding.On this basis,a fast video transcoding algorithm is proposed for H.264 noise video transcoding.The algorithm is divided into two parts:motion region detection and macroblock mode mapping.In the motion region detection section,the algorithm determine the motion of each frame violent area,the static area and between the two areas by using the decoded motion vector information.The motion region detection is mainly for the subsequent macroblock mode mapping part for reference.In the macroblock mode mapping section,there are mainly two kinds of pattern mapping methods,one is the original macroblock mode is intra mode mapping method,the other is the original macroblock mode is inter mode mapping method.The mode mapping method analyzes the influence of noise video on the macroblock mode selection before and after transcoding,so as to find the correlation between the macroblock mode before and after transcoding.We select the most possible set mode,then select the minimum rate-distortion cost of coding as the best mode,which can reduce the computational complexity and save the time for the traversal patterns.The codec used in the paper is derived from the code framework JM 18.4 provided by the international standard organization.The full-coding algorithm,full-mapping algorithm and the contrast algorithm are used to verify the experimental data.In the transcoding quality,our fast video transcoding method is close to the full-coding algorithm compared with the full-mapping algorithm and the contrast algorithm at the same bit rate.In the transcoding speed,the fast video transcoding method proposed in the paper is about 66.3%faster than the full-coding algorithm,and is improved by about 41.1%compared with the contrast algorithm at the same quantization step. |
PDF Full Text Request