Skip to Main Content
A fast two-step marker-controlled watershed image segmentation method in CIELAB color space is presented in this paper. We choose a number of seed points distributed nearly uniformly as the makers to perform the first marker watershed segmentation step, and obtain superpixels of the input image. These markers have the minimal gradient in a 3 ×3 neighborhood, which is able to avoid placing them at an edge and to reduce the chances of choosing a noise pixel. After superpixels segmentation, we do not adopt the traditional region merging strategies based on the different features of the adjacent regions, but cluster the superpixels in a 5-D space composed of Lab color vector and the position coordinates of the superpixels to resolve the over-segmentation problem, which saves a lot of computation time. Experiments on various types of images demonstrate that our algorithm is faster than many other segmentation algorithms and very suitable for real-time applications.