This article is dedicated to one of the basic problems in computer vision – image segmentation. Theimagesegmentationproblemisoneoftheoldestandhardestincomputervision. Theaimofthisworkistoanalyzeexistingresearchesof the segmentation algorithm concerning its speed of operation.The main methods of image segmentation,such as split-and-mergemethods, region-growingmethods, Markovfieldmodeling, edgedetectionandgraphpartitioningmethods, are considered. The article is focused on analysis of one of graph-based algoritms, the efficientgraph-basedimagesegmentationalgorithm, proposedbyP. Felzenszwalb.
Recognizing existing solutions and research of the segmentation algorithm, an assertion was made that the proposed improvements allow to increase the quality of segmentation, but performance of the algorithm becomes worse. The problem of improvement of image segmentation algorithms is actual and needs to be studied further.