Time Varying Volumetric Data Exploration Based On Dense Correspondence

The visualization of time-variant volumetric data is a challenge in scientific visualization.How to interactively mine the volumetric data by users and better understand the data has always been a key concern.Time-variant data exist in various fields,ranging from astronomical exploration to theoretical physics simulation to biochemistry research and so on.However,any research and solution on the issue of time-variant data will have great significance and effect.People often need to visualize the time-variant data in manner of context-consistent and tracking the feature changes over time,when they are in scientific research and production.These correspond to the two major themes of the time-variant data study:transfer function synthesis and feature tracking.In order to solve these two major issues,we propose a unified framework based on dense correspondence.Through our framework,users can deal with the transfer function design and feature tracking in a coherence manner.Specifically,we apply the divide and conquer ideas to segment time-variant data.The user only needs to perform operations on keyframes,then the framework can automatically generate context-consistent results on non-keyframes.In order to implement the framework,we have to build the correspondence between the volume data,so we propose an algorithm for voxel-level correspondence(dense correspondence):block match.In previous studies,compute the dense correspondence between two volume data is a very time-consuming thing that directly become a major obstacle to the use of dense correspondence to explore the time-variant data.The block match algorithm of this paper has the characteristics of fast,accurate,etc.,using GPU parallel acceleration,so the time-consuming is never been the bottleneck.Then,in the fields of time-variant data reaserch,this paper first introduced the conception of dense matching,which is a great innovation in our work.Further,we use the framework proposed in this paper to make evaluations of transfer function design and feature tracking.Experiments show that our framework can be more user-friendly for feature tracking and transfer function melding.And through experimental comparison,we conclude that effectiveness of our framework has reached state of art.In addition,we also analysis the parameters and performance involved in the framework.It shows that the framework has a good efficiency and scalability.