Allen Hirsh

Biophysicist and Digital Artist

Dr. Hirsh is a biophysicist in Silver Spring, MD, specializing in new chromatographic technology for protein separation — technology he co-invented with his colleague Dr. Latchezar Tsonev.

When he is not doing chromatographic research, he works with his brother, Gene Hirsh, a trained classical painter and digital artist residing in California, exploring one of the fundamental challenges of photography: how to increase the size of an image while retaining its quality. This problem has preoccupied him for the last 20 years.

Before the advent of digital images the magnification of color images inevitably blurred edges, increased the visibility of pigment grains and moved colors towards gray. With digital technology, certain trade-offs became available, but they provided greatly flawed solutions. One could retain edge strength and color saturation by pixel duplication, but the blocks of identical pixels created by such algorithms quickly form large stair-step structures utterly unacceptable in quality work. If the new pixels are produced by interpolation, even using bi-cubic information that extends beyond local pixels, edges rapidly lose sharpness and color saturation declines markedly.

Programs such as Genuine Fractals™ and its improved descendant Perfect Resize™ represent a major step forward. The strategy employed is to identify edges, then characterize the patterns adjacent to the edge by advanced Fourier analysis so as to blow-up not just the edge but the complex pattern in which the edge is embedded. It is a powerful method, but it suffers from two significant limitations. Since sampling errors in very small images are often characterized by strong edges, these errors will be faithfully reproduced. This means that images smaller than about 2 MB are not addressable and 10x enlargement is the recommended upper limit for all images. In addition, those areas of the image that are not identified as containing edges are still allowed to lose focus and move towards gray, losing the subtle texture of the original.

In exploring this challenge, Dr. Hirsh has developed a powerful, mathematical painting engine — one that uses certain algorithms to address aesthetic control of edges as an alternative approach to this problem. In essence, he uses bi-cubic algorithms to recapture weakened edges — a process generally thought to be too difficult to be practical. By also providing robust edge detection filters, Dr. Hirsh demonstrates how one can control texture in gradient regions - a method that works extremely well, even on very small images.