Images produced by assembling a large picture from a number of smaller rectangular photographs are called photomosaics. Investigators interested in photomosaics will find this paper of interest, as it is a preliminary attempt at empirically measuring the relative costs and effectiveness of photomosaic generating algorithms. The text is well written and easy to follow, even for those outside the field, where it may have implications for complexity theory, data compression, and copyrighting.

Two algorithms are described; both are then tested with a variety of tile sizes (tiles are the smaller photographs from which the larger is made). The effectiveness of an algorithm was measured in terms of similarity between the original image and the photomosaic; granularity of detail; and variety of the selected tiles. Cost was measured in terms of running time and the number of available tiles.

The results clearly reveal that one algorithm was unacceptably slow. They also less definitively indicated other relationships, including that minimum viewing distances increase proportionally to the square root of tile area, and roughly increase in proportion to the number of tiles.

Future results will be more convincing if the tests use more than one image, measure the visual acuity of several participants, address lighting and contrast issues, control tile quality, and minimize subjectivity. Unfortunately, the question of whether the results obtained with a rather amorphous image on an 8½" by 11" sheet, viewed at distances of up to 48 feet, can be extrapolated, for example, to unfamiliar billboards with text, viewed from perhaps a quarter of a mile, is not addressed.