Abstract

We have examined certain transformations, linear and nonlinear, which transform frame-grabbed black-and-white photographic images into artistic renditions of the original. Such transforms resemble sketches, sketch and wash renditions, or mosaics. Generally these transforms involve a large reduction in information of the original photographic image, and we have attempted to measure this reduction using coding algorithms. The reverse, increase in information when an artist creates a painting from a sketch, was also evaluated. Lastly, we also studied the importance of the information retained and discarded in the recognition of faces in a series of artistically transformed, frequency filtered images. We conclude that information necessary for recognizing faces (and many other objects) is broadly and redundantly distributed throughout the spatial frequency spectrum.

© 1986 Optical Society of America