Due to cost consideration, the quality of images captured from surveillance systems usually is poor. To restore the super-resolution of face images, this paper proposes to use Orthogonal Locality Preserving Projections (OLPP) to preserve the local structure of the face manifold and General Regression Neural Network (GRNN) to bridge the low-resolution and high-resolution faces. In the system, a face is divided into four blocks (forehead, eyes, nose, and mouth). The super-resolution process is applied on each block then combines them into a complete face. Comparing to existing methods, the proposed method has shown an improved and promising result.
Lecture Notes in Computer Science 7197, pp.253-262