Segmentation of Visual Images by Sequential Extracting Homogeneous Texture Areas

• DOI: 10.4236/jsip.2020.114005
• Publication Date: 2020-11-09T06:25:48Z
• Authors: Alexander Goltsev; Vladimir Gritsenko; Dušan Húsek
• Keywords: Texture (cosmology); Texture filtering; Texture compression; Region growing

The purpose of the research is to develop a universal algorithm for partial texture segmentation any visual images. main peculiarity proposed procedure extraction only homogeneous fine-grained segments present in At first, an initial seed point found largest and most segment image. This expanded using region growing method. Other image are extracted analogously turn. second stage, merging belonging same class performed. Then, detected input neural network with competitive layers which accomplishes more accurate delineation shapes segments. fully unsupervised, i.e., it does not use priori knowledge on either type textures or number results development realized as computer program tested series experiments that demonstrate its efficiency grayscale natural scenes.