Gradient vector and local distribution based volume visualization

Transfer function design is a crucial step in direct volume rendering process. In real world datasets, different tissues may share overlap ranges of scalar values which makes distinguishing them a difficult task. In this paper, we present a novel transfer function setting in statistical space to emphasize strong boundaries in the volume data and explore local neighbourhood's intensity distribution of all the voxels. By combing the information got, an opacity transfer function is obtained. Moreover, gradient vector is another important data metric to differentiate various structures yet not been paid much attention. A color scheme is designed in RGB color space to map gradient vector to r, g, b component to not only visualize different materials but also reveal scalar value's variation in the volume. We test our method on several volumetric data and experiment shows that mean-deviation scatter plot in statistical space could reveal distribution patterns of each voxel in the volume clearer and more compact compared with intensity-gradient magnitude scatter plot. Besides, our method could effectively visualize strong boundaries and the proposed color scheme could reveal scalar value's variations in different directions thus provide users with more deep insight into the data.