Numerical approaches in principal component analysis for face recognition using eigenimages

This paper presents a novel and feasible method of implementing the face recognition technique based on eigenfaces. The method is intuitive, simple to express in mathematical terms, and flexible. We create a database of images and train these faces using the eigenface method to recognize a given face in the database. Another case, where the input image is non-facial is identified using our reconstruction algorithm developed. We applied preprocessing algorithms like the smoothing transformation mean filtering, back ground elimination and local enhancement filter to bring the images in the database and probe image into a standard, recognizable format. Based on its ability to distinguish between different faces, the system showed a maximum recognition rate close to 90%. The relationship between recognition accuracy, scale and rotation was also investigated.