A new efficient watershed based color image segmentation approach

Abstract

Color image segmentation is an emerging topic in current image processing research. There exist different techniques for the same. Region based approach like watershed algorithm is one of them. But, watershed approach normally results in problems like over segmentation, noise, etc. In this paper, an efficient approach for the color image segmentation is proposed. Here, the input image is converted from RGB to HSV. Then, V channel is extracted from the converted image and normalized between 0 and 1. Otsu's thresholding is applied on the normalized image. The resultant image is then finally segmented with watershed algorithm. The result obtained from the proposed approach is found to be better in comparison to that obtained from the classical watershed algorithm.