Face Recognition using PCA and constrained images

The face recognition problem is receiving closer attention in the computer vision domain, especially now when contact tracing is a new normal, when monitoring of venues is a top security concerns, or when crime is increasingly becoming digitalized. On the other hand, the principal component analysis (PCA) approach, as a detection and recognition model, has been around for a while. However, evaluations of the accuracy of the PCA model when image inputs are blurred have been sidelined. We propose the design of a PCA inspired biometric personal identification model (PIM) that works with blurred images. The proposed model emphasizes extraction of data from a blurred image input before face recognition. Image restoration is the first step, where the blurredness of an image input is reversed by converting the blurred image inputs into Eigen faces. Euclidean distances are then used to achieve face recognition. Pre-processing also involves resizing of images. An experiment was administered to evaluate accuracy rate, false acceptance rate, and false rejection rate on the AT&T dataset. Results indicate that the PCA inspired PIM achieves a commendable accuracy rate. The achieved accuracy rate of 96% connote a good biometric PIM for this category of problems.