Gradient Estimation Based On Potential Function And Application In Volume Rendering

Direct volume rendering (DVR) is one of the kernel techniques and the main research aspect of volumetric data visualization. Since DVR techniques can generate high quality image without any help of geometry graph, and the result shows the whole information of volumetric data, it is widely used in the field of medicine, astrophysics, hydrokinetics, military, entertainment etc. But at the same time, DVR techniques have serious problems—complex algorithm and low speed. So how to accelerate algorithms in case of high quality result becomes the most important subject of limiting the development of DVR techniques.Based on the analysis of DVR theory, this paper researches and practices the Ray-Casting algorithm. Through digging out Ray-Casting principles and flows deeply, find that gradient of volumetric data is quite important to DVR as the parameter of Classification and Shading. So, this paper optimizes the flow of Ray-Casting algorithm firstly. Takes gradient estimation module as one step of volumetric data pretreatment, treats gradient vectors as one part of volumetric data and inputs of DVR flows. Thus not only decreases times of gradient estimation, accelerates rendering speed, but also provides convenience of implementing and improving gradient estimation algorithm. Secondly, this paper implements several gradient estimation methods in existence, and then improves them trough building a potential function. The existing gradient estimation algorithms can not get good balance between rendering speed and image quality, this article builds a cubic B-Spline function to estimate gradients, and gives procedure of formula building. At last, through the transverse comparison, it shows that new method of gradient estimation gets a better balance between rendering speed and image quality.