Several techniques for evaluating interpolation methods are described in the relevant literature, but none evaluate the quality of the rendering. The objective of this thesis is to present a method of objectively comparing the effects of various interpolation methods on the quality of the final rendering.
There are two experiments; both test interpolation methods on ray-casting functions. The first experiment renders four synthetic models. An exact image is rendered by computing ray samples directly from a mathematical model. The model also synthesizes a scene of voxels, which are then rendered with various interpolation methods. The interpolated image that comes closest to the exact image is considered superior. In particular, it was shown that cubic interpolation and linear interpolation are more accurate than nearest-neighbor interpolation. The second experiment renders actual MRI data.