Construction of a Corrector for Distorted Situations in Micro-object Images Based on Transformation and Filtering Operators

The scientific and methodological foundations were studied and methods, models, and algorithms for optimal identification of micro-objects were developed using Gaussian transformation and filtering operators, median filtering, fast Fourier transform, continuous and discrete wavelet transforms, and shift transforms. Mechanisms have been developed for using redundant information and structural components of an image based on the use of statistical and dynamic characteristics, which have advantages in reducing the complexity and labor intensity of information processing, identifying the redundant structural component, segmenting the image contour, growth dynamics, internal features and properties, visual delineation of approximations, and interpretation of objects. Computational schemes have been implemented that perform the functions of aligning histological sections, finding the contour of random situations and constituent objects, determining a set of levels and thresholds, performing segmentation, and conducting registration. Functional modules have been designed for graph search, image approximation based on wavelets, shear and other transformations, parameter determination, color coding and visualization of micro-objects. Algorithms for identifying, recognizing, and classifying microobjects were tested using real-world examples, specifically images of single-cell inflammatory cells (fibroblasts, fibrocytes) associated with lung disease. Signs of chronic inflammation, such as the presence of giant cells, were assessed. A software package for visualization, recognition, and classification of images of micro-objects has been developed and implemented. Its effectiveness has been studied under conditions of a priori insufficiency, parametric uncertainty, and low reliability of information.

Перевод пока недоступен