These patterns don’t include intensity variation and aren’t edge points and neural network returning white-white-white- white (Whitewash) for both. We can also reduce noisy pattern by Whitewash. Noisy patterns contain one black pixel that replace by Whitewash (see TABLE. 1). In Fig.2 output patterns also represented. After the network is trained, it can recognize the input pattern as a most like pattern in our edge pattern bank.