Valley based multiclass thresholding for color image segmentation
Tapaswini Pattnaik, Priyadarshi Kanungo · 2017
Colour image segmentation is an important and crucial step for any computer vision task. In real time application the segmentation method should be simple and fast. The most non trivial issue of any multi class segmentation problem is to estimate the number of levels in an image. In this paper we proposed a valley based thresholding method for color image segmentation. Our proposed method is capable to detect the number of levels through dominant peaks of each R, G, B color histogram. In terms of time complexity and average performance measure proposed method is better than the Otsu's multi class and JSEG based segmentation approaches.