Abstract: In order to solve the problem of incomplete edge of gray image by simple adaptive template in edge extraction by cellular neural network (CNN), a method is proposed to improve the simple template and refine the four-part template into six-part template. Firstly, the judgment range of the simple template is divided evenly. Secondly, the values of each adaptive template in different ranges are derived again. Finally, the edge extraction of the gray image commonly used in image processing is carried out and the results of the simple adaptive template are compared. Experimental results show that the improved adaptive template can extract more complete and clear edges than the simple adaptive template.