Recognizing Handwritten Numerals Using Multilayer Feed Forward Backpropagation Neural Network

Image processing is simply the processing of the given image. The input is just an image, that may be from any source, and the output may be an image or a set of parameters that are related to that particular image. Recognition plays an important role in the area of image processing. In this research work, the authors focus on the recognition of handwritten digits. A new method that uses neural network is proposed for recognizing handwritten numerals. It consists of three phases namely preprocessing, training and recognition. Preprocessing stage performs noise removal, binarization, labeling, rescaling and segmentation operations. Training stage adopts backpropagation with feedforward technique. Recognition stage recognizes input images of numerals. The proposed method is implemented in Matlab. The method recognizes the numerals with accuracy in the range 95-100%. It performs well and maintain the accuracy level even in case of deformed images and images of any size.