Gaussian Transfer Function Based on Boundary

In direct volume rendering, transfer functions map data points to optical properties such as color and opacity. This paper advances the use of boundary information in transfer functions with a methodology for computing high-quality boundary measurements. An efficient approach for the extraction of boundary from volume datasets was describes. First, boundary voxels are defined by gradient magnitude threshold to refine the dataset. And then an objective function to detect the good boundary can be defined as the surface integral of the mean gradient magnitude. The boundaries are detected as maxima in cumulative Laplacian-weighted gray value histograms divided by boundary area values, with a computationally efficient method that requires only a single pass through the boundary volume. Last, a Gaussian transfer function is designed by the boundaries for high-quality rendering, and the transfer function can be analytically integrated over a line segment under the assumption that data values vary linearly between two sampled points. Analytically integrated segment can then be composed using standard techniques.