A Watershed Segmentation Algorithm Based On Ridge Detection And Rapid Region Merging

The most important thing in metallurgic analysis is analyzing and measuring the grains segmented from the phase diagram,and the segmentation results have directly influence on the subsequent recognition and understanding.The metallographic image has the following features and properties,which are different from the ordinary image,such as disconnected boundaries,heterogeneous intensity,and internal grain blocks with blocking resistance and irregularity,and randomness of sample section.Compared to other image segmentation algorithms,the traditional watershed segmentation algorithm could engender better results,however,it resulted in over-segmentation meanwhile,with not so accurate results.In order to resolve the defects produced in the traditional watershed algorithm process,a watershed segmentation algorithm based on ridge detection and rapid region merging,was proposed in this study.In this new algorithm process,an extended terminal ridge method was presented to make up the discontinuous ridge.As the highest waterline in first watershed segmentation,ridge assisted the image segmentation,which is important basis in judging pseudo-blobs.Based on the proposed the method of extended terminal ridge this paper adjusts the mathematical formula,combined with the Bayesian Classification to judge the pseudo-blobs.Finally,using the method of the maximum similar region rapid merging to measure the similarity to direct the pseudo-blobs merging.The algorithm is tested in Matlab,and the results have shown that the new ridge detection method and the new merging rule could mark and merge the pseudo-blobs to the greatest extent,which solved the over-segmentation problem well,and produced good and satisfying results.