Adaptive perspective ray casting

Abstract

We present a method to accurately and efficiently perform perspective volumetric ray casting of uniform regular datasets, called Exponential-Region (ER) Perspective. Unlike previous methods which undersample, oversample, or approximate the data, our method near uniformly samples the data throughout the viewing volume. In addition, it gains algorithmic advantages from a regular sampling pattern and cache-coherent read access, making it an algorithm well suited for implementation on hardware architectures for volume rendering. We qualify the algorithm by its filtering characteristics and demonstrate its effectiveness by contrasting its antialiasing quality and timing with other perspective ray casting methods.