This paper presents cellular neural networks (CNNs), where output of a object gives edge detection. Firsts the output which is detected is used training set. Then the original picture is changed with noise. After training, CNN is applied to the changed picture. These templates which are taked from the first training are provided from the simple picture. Then the templates are used the other complex picture. Second, the output of the complex picture is used training set for the edge detection of the noisy picture. Last, the output of complex picture is prepared manually. This output is used training set then the edge of the complex picture is detected.