Digital image-tone restoration is examined in terms of information theoretic-based measurements. The metrics illustrate the relative degree of restorability of a 6-stop underexposure range on Kodak Tri-X pan film. An image chain model is developed to account for the discrete nature of sampled images used in digital image processing. The model includes the major system components (from object to viewed image) from which the system signal-tonoise ratio is calculated. Quantization noise and clipping are treated as signal-dependent, additive noise sources, and are shown to play a major role in degrading restored images. Although the eye characteristics are not included in the calculations, there is some indication that the logarithm of the information content and the noise equiva lent number of quanta correlate with the visual perception of image quality of a restored pictorial image; these metrics demonstrate their utility in characterizing the practical limits of an image-tone restoration system.
Library of Congress Subject Headings
Images, Photographic; Image processing--Digital techniques--Evaluation; Photography--Retouching--Technique--Evaluation
Department, Program, or Center
School of Photographic Arts and Sciences (CIAS)
Melnychuck, Paul W., "Information theoretic metrics for the assessment of digital image-tone restored images" (1984). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus