| Scientific and technological progress makes every day different from the previous day,which brings with it the emergence of massive video information.In the original video there is a huge amount of data,real-time video transmission,processing and storage are affected by bandwidth and hardware constraints,so the video compression coding technology has become the point which people care about.In recent years,the traditional video compression has been developing continuously.As people enhance the sense of security,surveillance video which involves more domains is becoming more and more widespread.How we can improve the efficiency for video surveillance coding,speed up transfer speed,reduce storage costs and the timely and effective information processing and analysis of surveillance video is a series of important chal enges we face.According to the features of the surveillance video,AVS monitoring extension level(AVS-S)has added a series of key technologies over the conventional coding framework.Background frame(G frame)not only has the basic function of I frame,but also is the only reference frame for background prediction frame(S frame).The background information in the surveillance scene that will be used by the S frame prediction is provided by the G frame.S frame has both the intra frame prediction encoding and the prediction of background frame.The combination of G frame and S frame can eliminate the redundancy in the background and effectively improve the efficiency of the monitoring video coding.In this paper,the updating method of G frame is analyzed and studied under the encoding framework of AVS-S,and a new method of G frame updating is proposed,namely iterative updating average method.Firstly,this paper presents four classical background modeling methods as the updating method of G frame by means of mean method,median method,segmented weighted running average method and Gaussian Mixture Model background modeling method.The simulation results showed that these four methods are all influenced by moving objects in the process of generating background models,and found that the modeling effects of the four methods are not good in the region where objects passed slowly.At the same time,the experimental results confirm the efficiency of video coding are influenced by the quality of G frame.This paper presents a new method of G frame updating,namely iterative updating average method.The method selects the pixel value with the highest frequency to determine the nature of each frame of the pixel in a certain number of training set,that is,the foreground pixel or the background pixel,according to high frequency of background area and low frequency of foreground area,discards the value of the foreground pixel,makes the average of the set of background pixel points as the final background pixel,and traverses all the pixels to obtain the final background model.In the classical background modeling methods,objects with slow moving speed are difficult to eliminate from the background model.And experiments prove that iterative updating average method overcomes this disadvantage to a certain extent,and better preserves the background information,and guarantees the quality of G frame,thereby further improves the coding efficiency of surveillance video.Under the encoding framework of AVS-S,the final coding efficiency of iterative updating average method is superior to segmented weighted running average method.Under the same code rate,the average increase of PSNR is 0.363 dB,and the average saving rate is 12.402% under the same PANR.In addition,the method is compared with the AVS baseline,and the average of PSNR is 1.186 d B under the same code rate,and the average saving rate is 43.283% in the same PSNR. |
PDF Full Text Request