Quadtree based color image segmentation method

Yu. A. Maniakov, A. I. Sorokin

Orel Branch of the Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 137 Moskovskoe Shosse, Orel 302025, Russian Federation

Abstract: The paper presents a color image segmentation method and an algorithm based on quadtree. The proposed method consists of several steps. First of them is border detection based on three-channel color and two masks. Then, the authors apply the thinning algorithm to decrease the area of the found boundary. The segmentation algorithm is divided into two parts. In the first part, the image is divided into segments as much as possible. In the second part, the segments union algorithm uses the finding neighbor's ID based on the FSM table and applies a link to ID to create a graph. The results of color images segmentation obtained on the basis of the described algorithm are presented.

Keywords: color image, segmentation, quadtree, edge, border, thinning, color reduction, split, merge, pixel.

Received: 14.09.2020

DOI: 10.14357/08696527200410