Abstract: In this paper we propose a novel frame work for face recognition in real-time environments using the Principal component Analysis (PCA)-based face recognition methodology. The proposed frame work is developed by three schemes namely, nonlinearity clustering, eigen vector mapping and relationship learning. In the beginning, a clustering algorithm is proposed as a preprocessing step. After clustering, the very low resolution (VLR), High resolution (HR), illuminated image (IL) pairs in every cluster is approximate nonlinear, i.e., the relationship will be approximately represented by a matrix. Then second resultant matrix spaces are converted into eigen vectors mapping for supporting nonlinear problem in real face images. Finally a kernel PCA model is used to learn relationship mapping, with completely different constraints. We develop a new information constraint is designed for human-based recognition and machine-based recognition to adopt real-time environment. The system proceeding the relationship learning between VLR images to the HR image space as well as IL image space for nonlinear problem. Based on learning map SR algorithm can reconstruct the image space and so measures the reconstruction error, rather than the existing algorithms that perform error the VLR space.