Application of the Model of a Spatially-Hierarchical Quadtree with Truncated Branches to Improve the Accuracy of Image Classification

Abstract

The article considers the problem of increasing the accuracy of the classification of satellite image samples under the assumption that their samples are independent. The increase in accuracy was achieved by additional processing of the image as a spatially hierarchical quadtree, which is a type of random Markov field. A modification of this model was proposed – a spatially hierarchical quadtree with truncated branches. For the original and modified models, we compared the classification results of a real radar image characterized by a large amount of noise. The classification accuracy was estimated as the proportion of correctly classified pixels within the selected homogeneous areas. It has been established that within the framework of the modified model, homogeneous areas of images are more correctly classified by transferring the properties of accumulated images of the same region to them. The modified model makes it possible to obtain a classification result of higher accuracy than the original one when processing noisy images, while having less resource intensity.