Chapter 7. MEASURES OF QUALITY DISCRETED IMAGES

When encoding images for storage or transfer using the minimum possible number of digits, it is required to keep the quality of the reproduced image within acceptable limits. To improve the visual quality of the image, image enhancement systems are being developed. Image correction systems are designed to compensate for distortions and create images that are close to those that would be obtained using an ideal system that does not introduce distortion. In all three tasks, there is a common question about image quality — about how to maintain, enhance, or restore it during image processing.

Image quality can be assessed in two ways: either by determining the fidelity of the reproduction, or by evaluating the decipherment of the image. The fidelity of reproduction characterizes the degree of deviation of the processed image from some reference, and deciphering indicates the extent to which a person or machine can extract useful information from the image. Most often, playback fidelity is associated with small differences between the processed and reference images. On the other hand, deciphering is, as a rule, due to the large differences between these images.

It is obvious that quantitative measures of fidelity and deciphering the image are very necessary for the design and evaluation of image reproduction systems. These measures will largely help to get rid of the time-consuming and sometimes inaccurate modern methods of image evaluation through subjective examination. In addition, on the basis of quantitative measures, methods for optimizing image processing systems can be developed.

Significant progress has been made in developing quantitative criteria for the fidelity and deciphering of images. However, the criteria introduced are not perfect enough: it is very often possible to give examples of images whose quality is formally assessed as high, and subjectively as low and vice versa. The creation of more sophisticated criteria for assessing the quality of images is undoubtedly connected with a deeper study of the properties of the human visual system.