Recognition of Human Faces

Abstract

This dissertation is devoted to the development of some image processing algorithms and implementation of these algorithms in face recognition system. At first face images were scanned and stored as windows bitmap file (bmp), in order to extract features. In features extraction, the original face, edge-detected face and threshold-face were taken to extract three types of features. The face was divided into small grids. Each grid of the face includes 10x10 matrix of gray level values. Thus 100 data were taken from a grid and the averages of these 100 data were taken as a feature. The recognition systems include the template matching by distance measurement (Hamming and Euclidean) between known and unknown features and a neural network. Fifty faces of ten persons were used to test the system. The highest recognition rate was 92%. Using features from threshold face and neural network as recognition tool achieved this result.