Abstract: Edge detection is an important task in Computer Vision for extracting meaningful information from digital images. The main goal of our proposed method is to obtain thin edges so that the result is more suitable for further application such as boundary detection, image segmentation, motion detection/estimation, texture analysis, object identification and so on. In this study, we propose a new approach based on gradient edge detection method (using canny edge detector) capable of operation on smooth edges. In this study we introduce convolution masks to obtain better edges. We have tested our results with Bonferroni test for mean performance and edge detector performance with Pratt measure and good acceptable results were obtained.