Blind Image Restoration Method Research Articles

Deep Variational Network Toward Blind Image Restoration.

Blind image restoration (IR) is a common yet challenging problem in computer vision. Classical model-based methods and recent deep learning (DL)-based methods represent two different methodologies for this problem, each with its own merits and drawbacks. In this paper, we propose a novel blind image restoration method, aiming to integrate both the advantages of them. Specifically, we construct a general Bayesian generative model for the blind IR, which explicitly depicts the degradation process. In this proposed model, a pixel-wise non-i.i.d. Gaussian distribution is employed to fit the image noise. It is with more flexibility than the simple i.i.d. Gaussian or Laplacian distributions as adopted in most of conventional methods, so as to handle more complicated noise types contained in the image degradation. To solve the model, we design a variational inference algorithm where all the expected posteriori distributions are parameterized as deep neural networks to increase their model capability. Notably, such an inference algorithm induces a unified framework to jointly deal with the tasks of degradation estimation and image restoration. Further, the degradation information estimated in the former task is utilized to guide the latter IR process. Experiments on two typical blind IR tasks, namely image denoising and super-resolution, demonstrate that the proposed method achieves superior performance over current state-of-the-arts. The source code is available at https://github.com/zsyOAOA/VIRNet.

重加权总变分结合hyper-Laplacian的图像盲复原方法

重加权总变分结合hyper-Laplacian的图像盲复原方法

Fourier ring correlation simplifies image restoration in fluorescence microscopy

Fourier ring correlation (FRC) has recently gained popularity among fluorescence microscopists as a straightforward and objective method to measure the effective image resolution. While the knowledge of the numeric resolution value is helpful in e.g., interpreting imaging results, much more practical use can be made of FRC analysis—in this article we propose blind image restoration methods enabled by it. We apply FRC to perform image de-noising by frequency domain filtering. We propose novel blind linear and non-linear image deconvolution methods that use FRC to estimate the effective point-spread-function, directly from the images. We show how FRC can be used as a powerful metric to observe the progress of iterative deconvolution. We also address two important limitations in FRC that may be of more general interest: how to make FRC work with single images (within certain practical limits) and with three-dimensional images with highly anisotropic resolution.

Joint blur kernel estimation and CNN for blind image restoration

Convolutional neural networks (CNN) have shown its excellent performance in computer vision fields. Recently, they are successfully applied to image restoration. This paper proposes a joint blur kernel estimation and CNN method for blind image restoration. The blur kernel estimation is based on both blur support parameter estimation and blur type identification. An automatic feature line detection algorithm is presented for blur support parameter estimation and a dictionary learning algorithm is presented for the blur type identification. Once the blur kernel estimate is obtained, we use an effective CNN for iterative non-blind deconvolution, which is able to automatically learn image priors. Compared with current blind image restoration methods, the proposed joint method can obtain restored images under three types of unknown blur kernels. The experimental result shows that the proposed blur kernel estimation algorithm can provide high accuracy results. Furthermore, the proposed joint blur kernel estimation and CNN algorithm is superior to conventional blind image restoration algorithms in terms of restoration quality and computation time.

Improvement of Robustness in Blind Image Restoration Method Using Failing Detection Process

Improvement of Robustness in Blind Image Restoration Method Using Failing Detection Process

Blind image restoration with sparse priori regularization for passive millimeter-wave images

Passive millimeter wave imaging often suffers from issues such as low resolution, noise, and blurring. In this study, a blind image restoration method for the passive millimeter-wave images (PMMW) is proposed. The purpose of the proposed method is to simultaneously solve the point spread function (PSF) and restoration image. In this method, the data fidelity item is constructed based on Gaussian noise assuming, and the regularization item is constructed as the hyper-Laplace function ||x||0.6, which is fitted according to the high-resolution PMMW images. Moreover, a data-selected matrix is proposed to select the regions that are helpful for estimating the accurate PSF. The proposed method has been applied to simulated and real PMMW image experiments. Comparative results demonstrate that the proposed method significantly outperforms the state-of-the-art deblurring methods on both qualitative and quantitative assessments. The proposed method improves the resolution of the PMMW image and makes it more preferable for object recognition.

Variational Dirichlet Blur Kernel Estimation.

Blind image deconvolution involves two key objectives: 1) latent image and 2) blur estimation. For latent image estimation, we propose a fast deconvolution algorithm, which uses an image prior of nondimensional Gaussianity measure to enforce sparsity and an undetermined boundary condition methodology to reduce boundary artifacts. For blur estimation, a linear inverse problem with normalization and nonnegative constraints must be solved. However, the normalization constraint is ignored in many blind image deblurring methods, mainly because it makes the problem less tractable. In this paper, we show that the normalization constraint can be very naturally incorporated into the estimation process by using a Dirichlet distribution to approximate the posterior distribution of the blur. Making use of variational Dirichlet approximation, we provide a blur posterior approximation that considers the uncertainty of the estimate and removes noise in the estimated kernel. Experiments with synthetic and real data demonstrate that the proposed method is very competitive to the state-of-the-art blind image restoration methods.

Guided filter-based blind image restoration method

As a major branch and research focus in the image processing field, the study on image restoration is always of practical significance and application value. Blind image restoration has always been one of the most difficult problems in image restoration. In view of the image motion blurring induced by the relative motion between the camera and the subject, we present a blind image restoration method based on guided filter. We first estimate the point spread function by using the iteration algorithm in the frequency domain. And then, because the guided filter has the edge-preserving smoothing property, we restore the target image by the non-blind image restoration algorithm based on the guided filter. By iterating the above two steps, we can obtain the original clear image. In order to verify the effectiveness of the proposed algorithm, we give several groups of experiments. Experimental results show that the proposed algorithm can not only effectively eliminate the noise and suppress the ringing, but also well preserve the edge and texture details. Therefore, the proposed algorithm can restore the image with high quality.

Effect of SNR Estimation in Defocus Blurred Image Restoration

A blind image restoration method is proposed to improve the quality of the image blurred by camera defocus and system noise. Firstly, the focus point spread function (PSF) of the blurred image is estimated through error-parameter analysis method. Secondly, the Signal-to-Noise Ratio (SNR) of the blurred image is estimated through local deviation method. Thirdly, utilizing the estimated defocus PSF and SNR, image restoration is performed through Wiener filtering method, in which circulation boundary method is adopted to reduce ringing effect. Experimental results show that the SNR of the blurred image is estimated approximately, and verify the great effect of SNR estimation in blind image restoration.

A Blind Restoration Method for Remote Sensing Images

This letter proposes a blind image restoration method for the deblurring of remote sensing images. A simple but robust identification method of point spread function (PSF) support is proposed, and a joint estimation method is presented to simultaneously solve the PSF coefficients and restoration image. To narrow the solution space for the best possible definition, the Huber-Markov (Huber-Markov random field) prior model is employed to regularize the two series of unknowns. Experiments were performed to demonstrate the effectiveness of the proposed approach.

Blind restoration of real turbulence-degraded image with complicated backgrounds using anisotropic regularization

This paper proposes a novel blind image restoration method based on estimating the point-spread functions by using two real turbulence-degraded images as input. The non-negative constraint and the spatial correlation are transformed mathematically into the penalty terms and added to the objective function. An anisotropic and nonlinear regularization function is proposed to adequately punish the differences of the point spread functions (PSFs) in the process of optimization estimation. Some definitions of weighted second-order differences are given and a fast method to construct the matrix of second-order weighted gradient operator is derived. The PSF values can be quickly estimated. With the estimated PSFs, the true images can be recovered by non-blind restoration methods. Experiment results for the restoration of real turbulence-degraded images with complicated backgrounds support the effectiveness of this proposed method.

Noniterative blind image restoration based on estimation of a significant class of point spread functions

Blind image restoration means estimating the true image from an observed degraded image, with incomplete or little prior knowledge of the blur function and the true image. Commonly, iterative or recursive methods are applied to this problem, but they are usually time-consuming, uncertain, and divergent. Focusing on a limited but significant class of point spread functions (heavy-tailed Levy processes), a noniterative blind image restoration method is proposed, which directly estimates the point spread functions from the degraded image's Fourier amplitude and restores the image by general slow evolution of a continuation boundary. This method is based on the fast Fourier transform, so it can accomplish the blind restoration of a 512 X 512 image in seconds of CPU time on a current desktop platform. From several experiments on synthetic blurred images and on actual images, it is found that the quality of the restored images has been greatly improved, and many fine details are restored, by applying this method. (c) 2007 Society of Photo-Optical Instrumentation Engineers.