The detection of spatially-varying blur without having any information aboutthe blur type is a challenging task. In this paper, we propose a noveleffective approach to address the blur detection problem from a single imagewithout requiring any knowledge about the blur type, level, or camera settings.Our approach computes blur detection maps based on a novel High-frequencymultiscale Fusion and Sort Transform (HiFST) of gradient magnitudes. Theevaluations of the proposed approach on a diverse set of blurry images withdifferent blur types, levels, and contents demonstrate that the proposedalgorithm performs favorably against the state-of-the-art methods qualitativelyand quantitatively.
展开▼
