Research On Ray Casting Algorithm Based On CT Image

Visualization in medical image is a hotspot in today's field of computer science, which can make up for the deficiency of imaging in medical imaging equipment, the technology supplies doctors with a realistic three dimensional images, and helps them observe, analyze from multi-level and multi-directional easily, so more and more scholars are concerned about it, and the technology gradually becomes a new and characteristic science.Visualization in medical image can be divided into two major categories: volume-based rendering and surface-based rendering. Ray casting is a typical space-image method in direct volume rendering algorithm, it is simple in principle, easy to implement and can generate high-quality images, but the drawing speed is slow. This article presents an accelerative method based on the crossing-area technique, OBRC for short.Re-sampling and calculation are key factors which influence the speed of rendering in ray casting algorithm, in addition to this, large numbers of lights are another important factors. Using bounding box to cut the volume data which is able to remove the empty volex around the three dimensional data set, and projecting the bounding box to imaging plane where emitting optical line from, as a result, it can decrease the amount of ray lines.The transfer function affects the image quality based on ray casting algorithm. One -dimensional transfer function is easy to design, but there is ambiguity; mutil-dimensional transfer function can generate high quality images, but its design complex and computing large. In this paper, based on one -dimensional transfer function, we add gradient information of volex so that it may characterize different matters, this method eliminates the ambiguity of one-dimensional transfer function and resolve the complexity of mutil-dimensional transfer function.Obtaining the scope of sampling points through intersecting between light and bounding box; Constructing a design of octree which is bale to optimize the selection and calculation of sample points, and neighborhood nodes of octree instead of equally spaced sampling points, so it avoids re-calculation to region in similar value of volex, moreover, the calculations of re-sampling do not need tri-linear interpolation, for this reason, it reduces the computational complexity.Experiment results show that the method based on octree proposed in this paper not onlt guarantees the image quality, but also meets the real-time rendering requirements.