Error Control of Identification and Filtering of Micro-Object Images

Abstract

Researched and developed scientifically and methodologically foundations for optimal identification of micro-objects using traditional and Gaussian filtering, median filter, filters based on fast Fourier transform, wavelet transforms, shift transforms, mechanisms using geometric, specific features, statistical, dynamic properties of image information. Mechanisms for optimizing the identification of micro-objects are proposed that have advantages in reducing the complexity and laboriousness of analyzing the structure and processing information, identifying and segmentation of the image contour, using the dynamics of growth, visual differentiation, extracting internal features and properties, approximation, smoothing, and interpretation of objects. A mechanism has been investigated and implemented that performs the following functions: aligns histology slices; finds contours of objects, a set of levels, thresholds, combines segmentation, conducts registrations, forms a search graph, performs approximations based on a wavelet, shear, and other transformations, determines parameters, performs color coding and color visualization of micro-objects. The implementations of algorithms and software modules of the software complex for identification, recognition and classification of micro-objects, in particular, cellular elements of the inflammatory series (fibroblasts, fibrocytes) of lung disease, have been tested. The signs of chronic inflammation were assessed - the presence of giant cells. A software package for visualization, recognition, classification of images of pollen grains has been developed, the implementations of which have been tested taking into account the conditions of a priori insufficiency, parametric uncertainty and nonstationarity of processes.