Natural images often suffer from common problem of poor resolution and low SNR. However, conventional methods can’t accurately detect the edge for this type of images. In this paper, we present a robust multi-scale edge detection algorithm for noisy images. Firstly, we down-sample the image to multi-scale resolution, and then the edge features are exacted with the improved difference eigenvalue algorithm. Finally, the multi-scale edges are combined to the accurate edge using the method of Improved New Edge Direction Interpolation (INEDI). Experimental results show that the proposed method outperforms the conventional methods while suppressing noise and preserving edge thus is suited for nature images edge detection.