The bubble column is a multiphase flow reactor, which has been widely used in energy and environmental field. The size and concentration of bubbles are of great significance for studying the heat and mass transfer process in the bubble column. A digital image processing technique is proposed to measure the bubble size distribution in the bubble column. However, the dense bubbles are easy to overlap during the bubble recognition process. To solve this problem, the overlapping bubble matching and the circumferential fitting based on the curvature calculation are presented. The segmentation and contour reconstruction algorithm are further used to determine the size distribution of bubbles. Experiments are implemented in a bubble column. Experimental results show that the proposed algorithm can not only extract the clear and complete bubbles from the image, but also accurately segment the overlapping bubbles of the image. The multi-scale bubble size distribution can be accurately obtained accordingly. As the gas flow rate increasing, the number of small bubbles increases sharply. At the same time, larger bubbles are also generated. The maximum diameter of the bubble and the average diameter of Sauter increase with the increase of the gas flow rate. But, their ratio remains basically the same. The configuration of the distributor has an influence on the uniformity of the bubble size distribution. The square distributor produces the most uniform bubble, and the gas holdup is higher than those of the other two types of distributors. These results verify the feasibility of the image segmentation and contour reconstruction method for the measurement of bubble parameters in gas-liquid two-phase flow.