Abstract

The discrete image restoration problem consists of solving a set of linear equations where g represents the measured image, the M × N matrix H denotes the degradation operator and f represents the image to be determined. Since H is usually an ill-conditioned or rank-deficient matrix, solving the set of equations is not straightforward. In addition, the set of linear equations are inconsistent due to measurement errors, quantization noise, and inaccuracies in modeling. Therefore, direct inversion techniques are not applicable.

© 1989 Optical Society of America