Abstract:Following the hierarchical Bayesian framework for blind deconvolution problems, in this paper, we propose the use of simultaneous autoregressions as prior distributions for both the image and blur, and gamma distributions for the unknown parameters (hyperparameters) of the priors and the image formation noise. We show how the gamma distributions on the unknown hyperparameters can be used to prevent the proposed blind deconvolution method from converging to undesirable image and blur estimates and also how thes… Show more
Image blur kernel classification and parameter estimation are critical for blind image deblurring. Current dominant approaches use handcrafted blur features that are optimized for a certain type of blur, which is not applicable in real blind deconvolution application when the Point Spread Function (PSF) of the blur is unknown. In this paper, a Two-stage system using Deep Belief Networks (TDBN) is proposed to first classify the blur type and then identify its parameters. To the best of our knowledge, this is the first time that Deep Belief Network (DBN) has been applied to the problem of blur analysis. In the blur type classification, our method attempts to identify the blur type from mixed input of various blurs with different parameters, rather than blur estimation based on the assumption of a single blur type in current methodology. To this aim, a semi-supervised DBN is trained to project the input samples in a discriminative feature space, and then classify those features. Moreover, in the parameter identification, the proposed edge detection on logarithm spectrum helps DBN to identify the blur parameters with very high accuracy. Experiments demonstrate the effectiveness of the proposed methods with better results compared to the state-of-the-art on the Berkeley segmentation dataset and the Pascal VOC 2007 dataset.
Image blur kernel classification and parameter estimation are critical for blind image deblurring. Current dominant approaches use handcrafted blur features that are optimized for a certain type of blur, which is not applicable in real blind deconvolution application when the Point Spread Function (PSF) of the blur is unknown. In this paper, a Two-stage system using Deep Belief Networks (TDBN) is proposed to first classify the blur type and then identify its parameters. To the best of our knowledge, this is the first time that Deep Belief Network (DBN) has been applied to the problem of blur analysis. In the blur type classification, our method attempts to identify the blur type from mixed input of various blurs with different parameters, rather than blur estimation based on the assumption of a single blur type in current methodology. To this aim, a semi-supervised DBN is trained to project the input samples in a discriminative feature space, and then classify those features. Moreover, in the parameter identification, the proposed edge detection on logarithm spectrum helps DBN to identify the blur parameters with very high accuracy. Experiments demonstrate the effectiveness of the proposed methods with better results compared to the state-of-the-art on the Berkeley segmentation dataset and the Pascal VOC 2007 dataset.
“…Most of the available blind deconvolution methods are iterative; see, e.g., [2, 3,5,6,7,10,12,14,15,16,18,19,21] and references therein. Recently, Justen and Ramlau [13] introduced a novel non-iterative method.…”
Blind deconvolution problems arise in many image restoration applications. Most available blind deconvolution methods are iterative. Recently, Justen and Ramlau proposed a novel non-iterative blind deconvolution method. The method was derived under the assumption of periodic boundary conditions. These boundary conditions may introduce oscillatory artifacts into the computed restoration. We describe extensions of the Justen-Ramlau method that allow the use of Neumann and antireflective boundary conditions.
“…Several methods tackle the deconvolution problem using the variational approach (see, for example, [5][6] [7][8]).…”
Section: Introductionmentioning
confidence: 99%
“…Using the variational framework we utilize a hierarchical Bayesian paradigm (see, for example, [7][9]) to jointly provide estimates of the posterior distributions of the restored image and the hyperparameters when a GGMRF prior is used. We develop two algorithms using our framework.…”
In this paper we propose novel algorithms for image restoration and parameter estimation with a Generalized Gaussian Markov Random Field prior utilizing variational distribution approximations. The restored image and the unknown hyperparameters for both the image prior and the image degradation noise are simultaneously estimated within a hierarchical Bayesian framework. We develop two algorithms resulting from this formulation which provide approximations to the posterior distributions of the latent variables. Experimental results are provided to demonstrate the performance of the algorithms.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.