Most-binarization method for restoration of linearly degraded binary images

A most-binarization method using a priori knowledge of binary-valued original images is proposed for restoration of linearly degraded binary images. Properties of the solution of this method are discussed, and an algorithm for the solution is offered. Meanwhile a sufficient condition of a nondegenerate LSI system is presented. Experimental results are compared with those obtained using constrained-least-squares and maximum-entropy methods. It is shown that the most-binarization method greatly improves the restoration quality.