This article proposes a light-based tranfer function to specify the illumination coefficient. In this transfer function, the X and Y directions are the scaler values of voxel sampled in two opposite directions in the gradient direction. To represent voxel in two opposite directions, the author proposes a new line-based histogram interface. The above line is the sacle value in one direction, and the following line is the scaler value, if the two points are in a straight line, they are in the same substance; otherwise, they are different substances. Specify the illumination coefficient, which is a one-dimensional transfer function based on the scale value. Specify diffuse, ambient, and specular respectively. Finally merged. I feel that the LH histogram in the tvcg previous article in is similar to its method. They both use the difference of two values to represent the gradient, rather than the pre-calculated gradient.
Finally, this article samples a hybrid solution to find the gradient. First, it is estimated that the gradient is stored. Since the density changes after the common opacity transfer function (this is the normal transfer function), the gradient direction of each voxle in volume may change. So the author finds the two Sclar values in the gradient direction, finds a new gradient, and compares it with the gradient obtained at the beginning, and selects a large gradient.
In, Eric Lum's pg paper proposed the multi-level transfer functin based on this new tranfer function interface, because this interface has obvious relationships and can be used. Because it mentioned in the lighting Transfer Function article that uses this interface as a commonly used transfer function, but it certainly won't be able to publish a paper. because I had to think about a new idea, I finally got the multi-scale tranfer Function Based on filter bank in.