On the Algorithm Optimization for Calculation a Stationary Flow on an Oriented Graph

Abstract

A method for obtaining the classification attribute of images based on the construction of a stationary flow on a graph constructed from a given image is considered. Weighted entropy, which is considered as a classification attribute, is calculated by the initial and stationary flows. Various approaches for optimizing this algorithm are considered in the paper. One of the proposed approaches is to divide the original image into a number of areas, the calculations for which can be performed independently from each other on different processor cores. The second approach is based on splitting the image into cells of a given size and constructing a graph whose vertices are associated with these cells. The third approach is to use in the data representation the so-called immutable collections that allow parallelization without synchronization. Comparative results of numerical experiments are presented.