The volume rendering technique for 3D display of organ surfaces from 3D data array was found to be superior to the other techniques in image quality. Surface-shading calculations are performed at every voxel with local gradient vectors serving as surface normals. In the separate step, nonbinary surface classification operators are applied to compute a partial opacity for every voxel. The resulting shading and opacities are then composites along viewing rays to form an image. Independence of shading and classification calculations ensure an undistorted visualization of 3D shape, and the nonbinary classification operators ensure that small or poorly defined features are not lost. but this technique requires large main memory and high computational expense.
In this thesis, we have implemented 3D surface display from volume data by using the computer system utilities efficiently, while maintaining the merits of the volume rendering technique. Computer memory assignments were reduced by reading and processing vowels slice by slice in main memory of the computer, and the processing time was reduced by applying front-to-back trancing method instead of back-to-front tracing method which needs redundant computation in voxel resampling along the rays. And we have proposed a new resampling method. In contrast to the conventional evenly-resampling interval, the proposed method reduced quantization artifacts of resulting image by resampling 3D color and opacity arrays by opacity-dependent interval.
