Image Segmentation Method Using Hybrid Region Flow

Abstract

This paper suggests an efficient image segmentation algorithm in 3 steps for grayscale images. First step is the segmentation as homogenous region on image. And, second step is merging of region without meaningful boundary between segmented regions. At the final step, region, which is considered as meaningless due to composition with minute region, is merged by consideration of similarity with surrounding regions. In this paper, 3 kinds of dams, energy dam, sponge dam, valve dam, are proposed to induce efficient merging for geometrical features. In addition, we propose the mask for merging of segmented region from gradation. Through this, we propose the efficient image segmentation algorithm which is robust from over-segmentation and over-merging.