In this paper, Wavelet-Network (WN) model has been recently proposed and applied to image processing, e.g., identification type of noise in Gray-Scale Images (GSI). This paper develops a new technique, which employs a Discrete Wavelet Transform (DWT) and an Artificial Neural Network (ANN). This WN technique uses special mother wavelet y(xl,x2) of (DWT) as activation function for (ANN) instead of the traditional activation functions like (Log sigmoid, Tan sigmoid, etc). It is shown here that the benefit of WN circuits which uses WN is a good approximation tool for GSI images. These approximation patterns for images forced ANN to learn on these images which will be used in the test phase after that.