In order to minimize the impact of anthropogenic factor and to study the relationships between dimple diameter and common mechanical properties,computer image processing technology is introduced to measure and count the size and number of dimples in metallographic image formed by ductile fracture.The dimples are segmented very well by using an adaptive method for image thresholding based on index of fuzziness.An improved multiregional area algorithm based on Freeman chain is presented.It proves that this algorithm can detect and measure dimple diameter rapidly and accurately.It also provides quantitative results for further classify and analysis of dimples.