The Separability Of The Blur Region In Single Image

Ubiquitous image blur brings out a practically important question – what are effective features to differentiate between blurred and unblurred image regions.Especially,it is more challenging in a single image when the blur is spatially-varying.Developing discriminative blur features is an open problem.In this thesis,we address it by studying a blur feature representations in image gradient.Specifically,we propose a new kernel-specific feature vector consisting of the information of a blur kernel and the information of an image patch,and the kernel specific-feature is composed of the multiplication of the variance of filtered kernel and the variance of filtered patch gradients.Unlike previous methods,which are often based on restoration mechanisms,our features are constructed to enhance discriminative power and are adaptive to various blur in images.The feature origins from a blur-classification theorem and its discrimination can also be intuitively explained.To make the kernel-specific features useful for real applications,we build a pool of kernels consisting of motion-blur kernels and defocus-blur(out-of-focus)kernels.By extracting such features followed by the SVM classifiers,the proposed algorithm outperforms the state-of-the-art blur detection method.Experimental results on public databases demonstrate the effectiveness of the proposed method.