Analysis and Experimental Research of Modifications of the Image Segmentation Method Using Graph Theory

In this work, we aim to research in increasing the processing speed of the image segmentation and improve the segmentation quality on textured images. We investigate the problem of image segmentation based on the color difference of regions. Then, we study the aspects of effective implementation of the image segmentation algorithm based on the minimum spanning tree graph. In particular, the use of different data structures for displaying segments. 

It is shown that the dependence of the segmentation result on the color difference metrics. It is suggested that modification algorithm with the use of an array of singly linked list and with the sorting of the graph edges over linear time, which has resulted in 4 times speed gain. We propose the modification algorithm with the use of superpixelization, which avoids the segmentation on the textured areas of the image achieved through the super pixel construction and its use as the graph nodes.