Defocus blur parameter estimation using polynomial expression and signature based methods

Restoration of images degraded by defocus blur can be done by using non-blind image restoration algorithms if the amount of defocus is known. This paper proposes two methods to estimate the radius of defocus in a blurred image. The first method establishes a mathematical relation between the defocus radius and the radius of the innermost concentric circle in the spectrum of the defocus blurred image. The second method uses signature to estimate the amount of defocus. Both the proposed methods are tested on standard datasets like USC-SIPI, Berkeley and Pascal VOC 2007. The experimental results are compared with existing techniques for radius estimation and they exhibit higher estimation accuracy for a wide range of defocus blur.