Visualization of random sampled data

The representation of scientific data as a visible surface using a natural scene paradigm is an attractive and intuitively meaningful way of visualizing data. Representation using this paradigm organizes the data in a manner which is closely related to the digital image; the data are represented as a scalar variable sampled over a two-dimensional spatial field. However, scientific data are seldom available in any uniformly sampled, regular form, thus requiring the use of some sort of modelling or interpolation scheme to produce the image-like representation of the data. The author discusses one approach to the visualization of randomly sampled data which produces an image in a single pass through the data, requires no high-resolution surface modelling, and suppresses moderate amounts of noise. The scheme is capable of interpolating between samples to fill in missing data regions. The technique is demonstrated using LIDAR data of coastal-region water depth. Data visualization using classical image processing methods is illustrated, once again using the LIDAR images as an example.