Study Of In-Situ Volume Rendering Based On Volume Depth Image

In-situ visualization is the most effective way to carry out PB and EB scientific data analysis.The key idea is to preprocess large-scale data in-situ and reduce the amount of subsequent transmission and storage,thereby alleviating storage and I/O stress.Different visualization tasks requires different in-situ processing methods.As to in-situ volume rendering,there are two problems mainly concerned:Perfectly embedding the volume rendering codes into the simulation calculation process&Achieving visual interactions in subsequent rendering as much as possible.First,this paper implements an in-situ volume rendering algorithm based on volume depth image.The algorithm is divided into two steps:Firstly,a modified ray casting algorithm is executed at the simulation nodes.The sampling points with similar color on one ray are combined into a super-segment to generate an intermediate data that stores depth and color information,which is called the volume depth image(VDI).Then the VDI is transmitted to the visualization nodes.At the visualization nodes,each super-segment in the volume depth image is extended to a quadrangular prism,and the quadrangular prisms are drawn to realize the volume rendering effects.The algorithm can successfully reduce the amount of data transmission from the simulation nodes to the visualization nodes,and it can correctly integrate the global VDI in the visualization nodes and then achieve high-quality,view-interactive visualization effects.Secondly,this paper presents a set of evaluation criteria for in-situ data processing,including:in-situ pre-processing time,in-situ data compression rate,and subsequent visualization quality.Combining the former algorithm,three evaluation values of volume depth image is given,including:VDI's generation time,VDI's data compression rate,and VDI's visualization quality.After analyzing the VDI's generation algorithm,the key parameter groups affecting the above three evaluation values are determined:the size of the VDI,the sampling interval when the VDI is generated,and the threshold value for judging whether two colors are similar.Finally,this paper proposes an auto-tuning framework for in-situ VDI parameters.A comprehensive and regulatable evaluation function for judging the quality of the volume depth image parameter group is given.Particle swarm optimization and genetic algorithm are used to speed up the parameter optimization process.Experiments show that this method can find the optimal parameter group quickly and reduce the workload of manual adjustment.