Resarch On Surveillance Videos Coding Based On Double Background Frame

Video surveillance system is a new application field of modern digital video technology,and it is a large information system,which contains of the developing video technology and internet technology.With the development of the society,the application of video monitoring system is becoming more and more extensive.It plays an important role in the fields of traffic,public security,analysis of big data and so on.However,massive video data poses a great challenge to both storage and transmission.Compared to the traditional natural video images,the surveillance videos have a fixed background,and a large number of background information will cause the background redundancy.Traditional video coding standards cannot meet the demand for surveillance video compression efficiency.Therefore,the research of coding technology for surveillance videos become one of the research hotspots at home and abroad.In view of the stable background of surveillance videos,the research of surveillance videos coding based on background prediction is carried out.The main contents are as follows: Firstly,a surveillance videos coding scheme based on dynamic background model is proposed by using the characteristics of fixed scene.The scheme improves the coding efficiency by building a background picture,which is used as a long-term reference in prediction.First of all,a dynamic background model based on coding block classification model is proposed.The first frame of video sequences is used to initialize the background picture,and block types are judged in block level.Then the background picture is updated with background region blocks dynamicly.In order to improve the accuracy of background updating,a background optimization algorithm based on blocks partition is proposed.In this method,the coding blocks are divided and blocks of different sizes are used to update background picture.Finally,the prediction coding scheme based on background picture is completed on HEVC encoding framework.The experimental results show that compared with the traditional coding standard,the proposed method can achieve overall 2.51% BD-rate gain.Second,in order to improve the quality of the background frame,a video coding algorithm based on double background frames is proposed.The main idea of the scheme is to train the original background frame with the original sequence,and the original background picture is used in the prediction process after the background encoding.The traditional background picture encoding methods are mostly based on intra coding,which results in a large number of bit rate increase while improving the quality of the background picture.The reconstructed background frame and the original background frame are utilized in double background frames coding scheme,and smaller quantization parameters are selected to ensure the quality of background picture.Experimental results show that the proposed scheme reduces the cost of background frames encoding while guaranteeing the quality of background frames.Meanwhile,compared with HM14.0,the compression rate of surveillance video has been improved by 28.21%.