De-aliasing Undersampled Volume Images for Visualization

Abstract

We present and illustrate a new technique, Image Correlation Supersampling (ICS), for resampling volume data that are undersampled in one dimension. The resulting data satisfies the sampling theorem, and, therefore, many visualization algorithms that assume the theorem is satisfied can be applied to the data. Without the supersampling the visualization algorithms create artifacts due to aliasing. The assumptions made in developing the algorithm are often satisfied by data that is undersampled temporally. Through this supersampling we can completely characterize phenomena with measurements at a coarser temporal sampling rate than would otherwise be necessary. This can save acquisition time and storage space, permit the study of faster phenomena, and allow their study without introducing aliasing artifacts. The resampling technique relies on a priori knowledge of the measured phenomenon, and applies, in particular, to scalar concentration measurements of fluid flow. Because of the characteristics of fluid flow, an image deformation that takes each slice image to the next can be used to calculate intermediate slice images at arbitrarily fine spacing. We determine the deformation with an automatic, multi-resolution algorithm.