Why don't we like blurry images?

Zhou Wang and Eero P. Simoncelli

Laboratory for Computational Vision, Center for Neural Science and Courant Institute of Mathematical Sciences, New York University, New York, NY 10003

Abstract: Humans are able to detect blurring of visual images, but the mechanism by which they do so is not known. A traditional view is that a blurred image looks “unnatural” because of the reduction in energy at high frequencies. We argue that the disruption of local phase is a more important factor for detecting blur. We first demonstrate that a blurred image that has its local phase corrected appears much sharper than one with its local amplitude corrected. We show that precisely localized features such as step edges result in strong cross-scale phase coherence in the complex wavelet transform domain, and blurring causes loss of such phase coherence. We propose a technique for coarse-to-fine phase prediction of wavelet coefficients, and observe empirically that (1) such predictions are highly effective in natural images, (2) phase coherence increases with the strength of image features, and (3) blurring disrupts the phase coherence relation in images. We thus lay the groundwork for a new theory of perceptual blur estimation, as well as a variety of algorithms for restoration and manipulation of photographic images.

Index Terms – blur, phase coherence, complex wavelet transform, steerable pyramid