Aggregate edge detection is the basis of creating concrete mesoscale model, which is applied to analyze concrete mesoscale characteristics. A concrete digital image edge detection method using DIS operator is presented in this paper. Mean filter, multi-scale filter, and Gaussian filter are compared on the effect of concrete image noise reduction. Based on the result, Gaussian filter is the most optimum method to reduce image noise and remain aggregate edge distinct. Sobel operator, Laplacian operator, and DIS operator are applied respectively to detect the aggregate edge on Gaussian filter processed images. Based on the experiment, DIS operator outperforms other two operators in the veracity and integrity of edge detection. It is concluded that using Gaussian filter and DIS operator for edge segmentation can provide geometrical models for FEM analysis.