Reconstruction of face image from security camera based on a measurement of space variant PSF

Images recorded by security camera are often severely degraded due to dirty lens or secular distortion of the recording system. To restore these images, the fully determination of space variant point spread function (PSF) is required. To measure PSF, we have proposed a method using a liquid crystal display, and shown some experimental results to restore the images by a CCD camera with intentionally distorted optics. The previous results, however, were not satisfactory due to the measurement error and noise contained in the measured PSF. In this paper, we propose a new method using M-array patterns for reduction of the measurement error and noise in PSF, and show improved results of restoration.