Quantitative analysis of marker-based watershed image segmentation

A methodology is proposed by combining the application of markers along with watershed transformation and thresholding for image segmentation. Use of the traditional watershed algorithm is widespread because of its advantage of being able to produce a complete division of the image. However, its drawbacks include over-segmentation and noise sensitivity. Therefore, the marker-based watershed segmentation is proposed here to overcome these effects. First, the original image is preprocessed by filtering techniques in order to smoothen it. Secondly, the foreground objects are marked. Then, the background markers are computed. Finally, the marked image is transformed through watershed transformation. The area is computed for the segmented objects in the image. It has been proved that this method reduces the error percentage.