Image processing involves modifying an image in such a manner that the objects of interest are accentuated to improve manual or automatic detection. It necessarily involves some information loss, so each algorithm to be applied must be considered carefully prior to invocation. Some typical image processing operations include noise removal, thresholding, smoothing, boundary detection, region labeling, and contrast enhancement. These and other operations are described in Jain (23).
Each of these operations has been the focus of significant research in the image processing community, although usually in the context of intensity or range images. Many of these concepts can be adapted to data of arbitrary dimension if it is assumed the data set is a packed (no gaps) discrete space; for sparse data sets empty space can be treated either as zero values or as a boundary into which new data may not be placed. Each operation helps to guide in the viewing of N-dimensional space by drawing attention to discontinuities or regions of uniformity and eliminating potentially insignificant and distracting data. Some of the techniques developed thus far in N-Land are described below.