Research On Video Streaming Based Cloud Game Coding

In recent years,with the continuous improvement of high-speed network and cloud computing technology,cloud gaming has gradually entered the public’s field of vision,arousing people’s great interest and attention.Compared with graphics streaming and hybrid streaming,the video streaming method has become the mainstream cloud game implementation solution due to its more stable data compression and lower performance requirements of game devices.However,the problems of large amount of data transmission and high bandwidth occupation are still important factors restricting the development of cloud game technology based on video streaming.Aiming at the challenges of huge network transmission and insufficient coding efficiency in the video streaming method,this paper implements a cloud game system based on video streaming coding optimization,and proposes two algorithms that can effectively improve the video coding performance.Traditional video codec algorithms can estimate and compensate for translational motion in video sequences well,but are not sensitive to the rotation and scaling of video images caused by camera rotation,forward and backward motion.In order to comprehensively improve the coding efficiency of cloud game video under different camera motions,this paper proposes an optimization algorithm for cloud game coding based on Depth Image Based Rendering(DIBR).The algorithm first generates a reference frame image with a higher degree of matching with the current frame through 3D warping and image coloring.Then according to the image,three reference frame list generation modes are proposed to correct the reference frame list.The experimental results show that the algorithm enables the encoder to estimate and compensate for camera rotation,forward or backward motions,and effectively improves the coding performance of cloud games based on video streams.When using the DIBR algorithm to generate a new reference image,the corresponding depth map needs to be transmitted through the network,which will result in additional transmission costs.Combined with the DIBR algorithm,this paper proposes an optimization strategy for cloud game coding based on depth estimation and multi-view blending.The depth estimation module first matches pixels between multiple views,then calculates the matching errors of a pixel under different depth labels,and uses the depth label corresponding to the minimum matching error as the depth estimation value of the pixel,thereby avoiding the coding and transmission overhead of the depth map.The multi-view blending module first analyzes the factors that affect the image quality,and then uses other reference frame information in the reference list to blend and output the final image,thereby further reducing the coding residual and the transmission data.On the basis of the above algorithm,this paper also designs and implements a cloud game system optimized for video stream encoding.The system consists of three parts: the engine,the server and the client.The engine renders the entire game scene and obtains information such as video frames,depth maps,and game camera data.The server is responsible for 3D Warping,video encoding,etc.,while the client completes video decoding and display.The experimental results show that the system can effectively reduce the video transmission bit rate and improve the video coding performance to a certain extent.