Research On Visual Human Volume Rendering Based On Mapreduce

With the increasing of medical volume data, the rendering speed has become amajor bottleneck hinder for the development of the volume rendering application, inwhich to ensure the draw image quality at the same time.The improvement of volume rendering algorithm, hardware performance andparallel technology are used to improve rendering speed. It is a viable way forimproving the algorithm of volume rendering. Hardware, now high performancecomputers have become possible, but when the data is large in scale and complexstructures, the speed of upgrade is still limited. With the development of networktechnology, distributed and parallel computing is applicated more and more, so usingparallel computing is a feasible method to improve the speed of volume rendering.The aim of MapReduce is problem decomposing which a large-scale dataprocessing breaks down multiple small scale datas processing. a plurality of smallscale datas are resolved by many computing nodes for parallel processing.In this paper, with MapReduce model, the traditional single volume rendering isintroduced into the distributed parallel computing environment.The method is toimprove rendering speed and realize the processing of massive medical data, whichcan solve the calculation of large volume rendering and the speed of image generation.The main contents of this paper are as follows:(1) Something about volume rendering are discussed in detail, especially theprinciple of ray-casting algorithm. The problems about the present volume renderingare described, as well as the feasibility for using MapReduce model to improve therendering speed in this paper.(2) The method has been put forward for the task partitioning based onMapReduce pattern drawing. A single host is not related to the task partition on thetraditional volume rendering, whitch the processing object is the whole body of datafield. The volume rendering under MapReduce should has a need for volumerendering task partition. The datas is divided into a plurality of small sub volumedatas, and then be submitted to multiple computing nodes for parallel processing.Each node is responsible for that part of the data belongs to own, and finally the rendering results of computing nodes are composited.(3) The ray casting algorithm has been improved. After improved Ray castingalgorithm, the draw processing of each nodes is parallel, the work node is responsiblefor processing sub data slicing cluster corresponding to the resampling and imagesynthesis, and send the results of rendering to master node. Master node completerendering results collection, generate the final image and display it.(4) The system has been designed for the virtual human rendering based onMapReduce. It is used to study virtual human volume rendering,whitch is designed toaccelerate the speed of rendering.Some test examples has been given on the system,the experimental results have shown that the visual human volume rendering based onMapReduce has an obvious promotion effect on the rendering rate.