Watershed: A Comparative Study

摘要

Morphological gradient is often used to find the gradient of an image which is further used for the transformation. However, noise in the gradient image results in to over-segmentation which have an undesirable bad effect on resulting segmented image. For the fine segmentation results the quality of the gradient estimate has a major influence on the segmentation performance. In this paper it is shown that different types of gradient are computed with the help of different operators and further these various gradients are used to segment the image with the help of watershed transformation. The resulting image constitute of watershed divide lines and the catchment basins. The purpose of this work is to show that the gradient obtained through different operators like Sobel, Prewitt, kirsch etc. have varying effect on watershed in terms of peak signal to noise ratio.