Research On Compressed Multi-Attribute Fused Volume Rendering In Higher Order Tensor

With the increasing scale of data,compressed volume rendering is an effective way to solve the issues of large-scaled volume rendering.The compression method based on tensor approximation has a good performance in terms of compression rate and rendering effect among the compression methods.Multi-attribute fusion is the mean to visualized data with reducing multiplicity and highlighting features.In this thesis,the research focused on multi-attribute compression volume rendering is conducted and achieves some expected results.It has great promising value both in theory and application.Aiming at the problem of compressed multi-attribute fused volume rendering,this thesis explores the fourth-order tensor approximation compression method based on the existing multi-attribute fusion by the third-order tensor approximation.Rank truncation in tensor approximation affecting the compression rate and performance obviously is also conducted in the research.The main work and innovation are as follows:1.A method of compressed multi-attribute fused volume rendering using the fourth-order tensor approximation is proposed.There are no effective methods and methods for the problem of multi-attribute compression fusion rendering.Yet the approximation based on 3rd tensor can be used to compact the multi-attribute data respectively and the fusion is implented in the rendering pipeline.This method has low compression and is not able to take advantage of the correlation among attribute.Thus a method based on the fourth-order tensor approximation is proposed.The basic idea is to take the data as a entirety in 4th dimension form and compact and fuse using 4th order approximation.The simulation results show that the method has obvious improvement in the compression rate of the data in the case of very little loss of data information;2.A method of higher order tensor approximation based on anisotropic rank truncation is proposed.The uniform rank truncation is usually used in all directions in approximation,while the actual data tends to be anisotropic in different directions.Therefore,a high-order tensor approximation method based on anisotropic rank truncation is proposed.To improve the compression performances,the rank truncation value is calculated in each direction by analyzing the singular value distribution in each direction,and the truncated ranks are obtained in different directions.The simulation results show that the method promotes the compact quality while satisfying the compression ratio.In summary,this thesis implements the above methods on simulation platform.The simulation results show that the proposed method improves the compression ratio of the data in the multi-attribute data compression fusion.And the methods proposed in this thesis are effective to improve the compression quality of data under different compression ratio conditions.