The application of watershed and region merging for image segmentation.

Abstract

Image segmentation is an important stage in doing image analysis. Among many different techniques in segmenting image, watershed can be noted attracted the attention of the researchers of this field in the past decade. In this paper, we use watershed transform to handle initial segmentation in our system. The vulnerability of watershed from the appearance of noise contained in the image is compensated by the application of region merging mechanism, in which we introduce the merging process relies on the calculation of integer value derived from each region intensity mean. Here we put more intention on merging small regions than big regions by assigning a different merging factor for both regions. From the experimental stage we will show the reduction that can be taken by the system in compromising over-segmentation produced by watershed transform.