On Image Evaluation and Enhancement

Abstract

Many techniques have been applied in image enhancement. Some of these are briefly reviewed here. Note that many of them may be classed under the general heading of ‘linear filtering’. To evaluate any image enhancement procedure, quantitative image assessment criteria are required. The criteria applicable to systems meant to convey information are reviewed here. It is noted that large areas of application can be covered by two, relatively simple, criteria: signal to noise ratio and mean squared deviation. Image optimization is investigated from the point of view of least squared deviation with both linear filters and scale change considered. The latter is treated in some detail. The term ‘image restoration’ is proposed to cover efforts to minimize the squared deviation of the actual image from its ideal version.