Abstract: This paper is an attempt to develop an image segmentation algorithm using the marker controlled watershed algorithm. The initial low-level segmentation step is considered as the major need for the fine segmentation. This low level segmentation must be of lower computational complexity and of lower time consuming. The watershed is the movement of the water from the catchment area to the place where water sources meet. The watershed in the image is similar to the catchment basin of a height map. The watershed lines are the defined on edges, nodes and also in the hybrid lines which lie on both nodes and edges. A spatially regularized gradient is introduced in order to a tunabletrade-off between the regularity and the adherence to the object boundaries. Matlab based implementation is carried out and the results are tabulated and the results are compared with the previous implementation.

Keywords: Super pixels, watershed, segmentation.