Adaptive Sampling Volume Rendering Algorithm That Based On Importance Of Data Space

Volume rendering is an effective data visualization technology, compares to traditional surface rendering, it does not need to create intermediate geometric primitives, but map the volume data to the two-dimensional images directly, which can describe the internal details effectively. Therefore, scientific research group and technical personnel pay more attention to it. Ray casting algorithm is a high image quality volume rendering algorithm and widely used, but high time complexity of the algorithms and slow rendering performance become the bottleneck to its further use. With the rapid development of graphics devices, in particular the introduction of GPU, this provides a new resolution of the issue.At present, adaptive sampling method that used by the ray casting algorithm is mainly used in empty space skipping, fast isosurface extraction and multi-resolution display. From the view of purpose, the first two methods reduce the algorithm time from the whole part, without affecting image quality. Therefore, they did not make the image quality improved; the latter one, from the view of users, reduces resolution of uninterested region to short algorithm time. Therefore, it does not improve the region that users interested. In addition, the sampling method does not base on volume data specifically, and disjointed from spatial data character, that ignored the spatial importance of volume data.Combining with the trend of volume data space, aim at the problem of Ray casting algorithm, the paper gives rendering algorithms that based on volume data: based on point of gradient and block importance that fully adaptive sampling rendering algorithm, and got a better rendering quality. At the same time, aimed at sampling algorithm of block importance, gave the spatial index structure that based on octree, achieved the fragment-level empty space skipping, and improved rendering speed. Finally, integration of two kinds of adaptive sampling techniques, paper designed and implemented the volume rendering system framework. The major work and research results are showed:(1) Giving a fully adaptive sampling rendering algorithm that based on gradient point, it solved visualization of volume data changes effectively. To improve the efficiency, the algorithm used pre-integration techniques in the synthesis stage, thus speeding up the sampling point conversion and pixel color synthesis. Meanwhile, algorithm avoids unnecessary data reconstruction and synthesis of computing, using compact bounding box technology, realizing empty space skipping to further improve the efficiency of the algorithm. Volume Rendering algorithm for the gradient calculation is completed in preprocessing phase, obtained the sampling point gradient by interpolation in the rendering phase, the dynamic calculation of the sampling step, breaking the independence between the sampling step and the characteristics of volume data. Experimental results show the material characteristics of contact surfaces clear, clear boundaries, a more detailed image, to retain the true space volume data ,detail the data significantly, enhance the body boundary of different material characteristics, highlight the space importance of the volume data, achieved good rendering results with the adaptive sampling of gradient point rendering algorithm.(2) A block-based importance of adaptive rendering algorithm is presented, which is improving on the point gradient algorithm, and the concept of the block importance is proposed. The main features of block importance sampling algorithm are: transform the sampling step that need to calculate in the fragment programs of point gradient, into only calculated the sampling points of data blocks and light intersect. The algorithm pre-calculated unit gradient of volume data and divide the volume data into multiple data blocks using octree. According the unit number where the node of octree leaves coverage and the unit gradient value of the block importance set the block’s sampling frequency, to achieve adaptive sampling. To improve the efficiency, an improved spatial index structure based on octree is proposed. The volume data based on octree organization can implement ray casting process in the fragment-level,which removed empty voxel on dynamicly. Experiments and analysis show that combine sampling methods based on block gradient and spatial index structure increased re-sampling efficiency compared with the traditional method; and not reduce image quality compared with point gradient algorithm, while improving the rendering speed of 25%.(3) In accordance with modern software engineering principles, we build visualization frame based on gradient points and blocks importance adaptive algorithm, give the class structure diagrams and sequence diagrams, clear the module’s function, clarify the relationship between modules. It shows good function of the framework scalability and reusability, through optimize the frame structures to further improve the operating efficiency.