(A) fast volume rendering algorithm using limited size of memory

Abstract

This thesis proposes a direct volume rendering scheme using 3D image data. Because the general volume rendering algorithm handles very big size of data and requires a lot of memory, one of the most serious bottleneck is the system memory size. The proposed algorithm utilizes Sliding Buffer technique to resolve this problem. Because small size of Sliding Buffer rarely depends on hard swapping when rendering procedure is performed, it is very efficiently performed in the systems equipped with small size of memory. Additional speedup is obtained by reordering the direct rendering procedure. Instead of resampling both of color and opacity, only resampling of color is needed since it is possible to perform the classification step following the resampling step. During the ray tracing step, each scan-line is encoded by run length coding after each voxel is thresholded to transparent or not, and the transparent ones are skipped at volumetric compositing step. This method makes a great amount of reduction of redundancy in volume data set. The proposed method thus works with great efficiency specially under the circumstances of using limited size of system memory.