Watershed Segmentation Using Curvelet and Morphological Filtering

An improved segmentation method is proposed in this paper for metallographic images, especially those objects surrounded with complex texture. If only watershed algorithm is used for the segmentation of an image, the over-segmentation problem will be serious. To solve this, we proposed a new approach. In the method, we take the curvelet transform to denoise initial images by thresholding the different scales of coefficients firstly. Afterward the improved morphological Top-Bottom filter is used for filtering the produced images, and then watershed algorithm is applied lastly. After segmenting, we do some simple progresses to remove some separated holes. The results demonstrate that combining curvelet and improved top-bottom filter can help us to get more accurate segmentation.