Abstract: In the research on the project onto convex set(POCS) super-resolution image restoration algorithm, it is observed that the edge region does not need the same relaxation factor as the uniform region. Revising this problem can reduce the large computational complexity of POCS and improve its practical application quality. Consequently, a fast POCS (FPOCS) is proposed. By quoting fuzzy entropy in the POCS image restoration process for edge detection, a monotonous increasing function that defines the relaxation factor is constructed based on the neighborhood homogeneous measurement (NHM). Therefore, the proposed approach can select the relaxation factor adaptively by the local character of the image. Experimental results show that the new approach can achieve a similar or even better restoration performance after dozens of iterations while the traditional POCS needs hundreds of iterations.

Key words: image restoration, super-resolution, fuzzy entropy, project onto convex set