METHOD OF PRE-PROCESSING TOMOGRAFIC IMAGES OF THE FUNDUS

Abstract

The paper deals with a number of modern methods and algorithms for image processing, namely: Kirsch filter, nonlinear filters Robert and Sobel, Wallace and SUSAN methods, as well as software for application of these methods on the example of analysis of tomographic images of macular changes of the retina. The most informative for image segmentation is the Kirsch filter based method and the Sobel nonlinear filter method. In contrast to the so-called deterministic distortions, which are often described by elemental functional transformations of the input image, models of additive, impulse and multiplicative noise are used to describe the random effects. As a rule, Kirsch, Roberts, Sobel algorithms are used to improve the quality of biomedical image processing. Features of construction of high-performance expert systems and devices of image processing and recognition by creation of neural-like methods and algorithms of parallel-hierarchical transformation, network methods of image pre-processing, development and creation of hardware specialized image computers based on them, which are combined with the eye, are considered. video processing (for example, the use of PCF, Q-conversion, parallel-hierarchical image encoding method, space-time ve matrix model of computing processes, parallel algorithms and structures that they reflect for parallel imaging) is a fairly urgent task. Thus, in contrast to the so-called deterministic distortions, which are often described by elemental functional transformations of the input image, it is advisable to use additive, pulse, and multiplicative noise models to describe the random effects. It is shown that, as a rule, the Kirsch, Roberts, Sobel algorithms are used to improve the quality of biomedical image processing.