Letter to the Editor

sir, a recent paper by Smith et al. (1999) in this journal demonstrated the utility of anaglyph stereo images for viewing surface range images such as those produced by the atomic force microscope (AFM). We feel that it is also useful to point out some additional ways in which stereo views can be employed to enhance the visualization of such surface images. Images in which the pixel grey scale value encodes elevation information (variously called surface images, range images or elevation maps) are produced by many techniques, including the use of AFM and its cousins, scanned stylus metrology instruments, interference microscopes, and even side-scanned radar and structured light techniques. Human visual recognition is not very sensitive to small changes in grey scale intensity and so we have difficulty in interpreting these images when viewed in their raw form. Methods such as contour mapping and pseudocolour rendition are sometimes used as an aid but tend to break up the overall gestalt of the image even as they reveal local variations. Simple derivative images, also widely used, tend to obscure the actual elevation information and hide details in some orientations. Surface rendering provides a tool that presents these surfaces as a human expects surfaces to appear, with lighting and surface specularity that can be adjusted to vary the appearance in useful ways (Bui-Tuong, 1975; Foley & Van Dam, 1984) (Fig. 1). Computing a red±cyan anaglyph image in which disparity is proportional to elevation produces a stereo pair view that is also helpful in many cases, and can be made more so in combination with rendering (Fig. 2). However, the point of view is still restricted to the surface normal, whereas in examining real surfaces people are able to orientate the sample to view it from various directions. We find it helpful therefore to combine surface rendering, colour coding and stereo viewing possibilities with perspective-corrected isometric views of the surfaces (Fig. 3). These present very realistic images that enable the human vision system to extract key morphological parameters and recognize features on surfaces. All of the calculations for these operations are well known, based on analytical geometry and rendering methods widely used in computer-aided design software, and have been implemented in various software packages. Our contribution has been to write the algorithms as a set of plug-in modules that can be used with Adobe Photoshope and compatible programs (which include NIH-Image, Image-Pro Plus and other