Research of metallographic image segmentation algorithm

Abstract

To solve such problems as low contrast and irregular particles in metal image segmentation and concerning the specificity of the images,an object segmentation method using convex point and the shortest distance was proposed. In this method,the kernel objects were separated from background by typical object thresholding,then noise was removed by mathematic morphology,edge smoothing,edge and concave point finding,cutting points matching and cutting line determination with corresponding of concave points,back extended lookup and the shortest distance. This algorithm was implemented and tested with metal image. The result shows that the image is effectively segmented without over-segmentation.