Blind Image Blur Estimation via Deep Learning

Estimation of non-uniform motion blur using a patch-based regression convolutional neural network

The non-uniform blur of atmospheric turbulence can be modeled as a superposition of linear motion blur kernels at a patch level. We propose a regression convolutional neural network (CNN) to predict angle and length of a linear motion blur kernel for varying sized patches. We analyze the robustness of the network for different patch sizes and the performance of the network in regions where the characteristics of the blur are transitioning. Alternating patch sizes per epoch in training, we find coefficient of determination scores across a range of patch sizes of R 2>0.78 for length and R 2>0.94 for angle prediction. We find that blur predictions in regions overlapping two blur characteristics transition between the two characteristics as overlap changes. These results validate the use of such a network for prediction of non-uniform blur characteristics at a patch level.

Learning local depth regression from defocus blur by soft-assignment encoding

We present a novel, to the best of our knowledge, patch-based approach for depth regression from defocus blur. Most state-of-the-art methods for depth from defocus (DFD) use a patch classification approach among a set of potential defocus blurs related to a depth, which induces errors due to the continuous variation of the depth. Here, we propose to adapt a simple classification model using a soft-assignment encoding of the true depth into a membership probability vector during training and a regression scale to predict intermediate depth values. Our method uses no blur model or scene model; it only requires a training dataset of image patches (either raw, gray scale, or RGB) and their corresponding depth label. We show that our method outperforms both classification and direct regression on simulated images from structured or natural texture datasets, and on raw real data having optical aberrations from an active DFD experiment.

Deep Ranking Exemplar-Based Dynamic Scene Deblurring

Dynamic scene deblurring is a challenging problem as it is difficult to be modeled mathematically. Benefiting from the deep convolutional neural networks, this problem has been significantly advanced by the end-to-end network architectures. However, the success of these methods is mainly due to simply stacking network layers. In addition, the methods based on the end-to-end network architectures usually estimate latent images in a regression way which does not preserve the structural details. In this paper, we propose an exemplar-based method to solve dynamic scene deblurring problem. To explore the properties of the exemplars, we propose a siamese encoder network and a shallow encoder network to respectively extract input features and exemplar features and then develop a rank module to explore useful features for better blur removing, where the rank modules are applied to the last three layers of encoder, respectively. The proposed method can be further extended to the way of multi-scale, which enables to recover more texture from the exemplar. Extensive experiments show that our method achieves significant improvements in both quantitative and qualitative evaluations.

DeFusionNET: Defocus Blur Detection via Recurrently Fusing and Refining Discriminative Multi-Scale Deep Features

Albeit great success has been achieved in image defocus blur detection, there are still several unsolved challenges, e.g., interference of background clutter, scale sensitivity and missing boundary details of blur regions. To deal with these issues, we propose a deep neural network which recurrently fuses and refines multi-scale deep features (DeFusionNet) for defocus blur detection. We first fuse the features from different layers of FCN as shallow features and semantic features, respectively. Then, the fused shallow features are propagated to deep layers for refining the details of detected defocus blur regions, and the fused semantic features are propagated to shallow layers to assist in better locating blur regions. The fusion and refinement are carried out recurrently. In order to narrow the gap between low-level and high-level features, we embed a feature adaptation module before feature propagating to exploit the complementary information as well as reduce the contradictory response of different feature layers. Since different feature channels are with different extents of discrimination for detecting blur regions, we design a channel attention module to select discriminative features for feature refinement. Finally, the output of each layer at last recurrent step are fused to obtain the final result. We collect a new dataset consists of various challenging images and their pixel-wise annotations for promoting further study. Extensive experiments on two commonly used datasets and our newly collected one are conducted to demonstrate both the efficacy and efficiency of DeFusionNet.

Defocus Image Deblurring Network With Defocus Map Estimation as Auxiliary Task

Different from the object motion blur, the defocus blur is caused by the limitation of the cameras' depth of field. The defocus amount can be characterized by the parameter of point spread function and thus forms a defocus map. In this paper, we propose a new network architecture called Defocus Image Deblurring Auxiliary Learning Net (DID-ANet), which is specifically designed for single image defocus deblurring by using defocus map estimation as auxiliary task to improve the deblurring result. To facilitate the training of the network, we build a novel and large-scale dataset for single image defocus deblurring, which contains the defocus images, the defocus maps and the all-sharp images. To the best of our knowledge, the new dataset is the first large-scale defocus deblurring dataset for training deep networks. Moreover, the experimental results demonstrate that the proposed DID-ANet outperforms the state-of-the-art methods for both tasks of defocus image deblurring and defocus map estimation, both quantitatively and qualitatively. The dataset, code, and model is available on GitHub: https://github.com/xytmhy/DID-ANet-Defocus-Deblurring.

Using a Blur Metric to Estimate Linear Motion Blur Parameters

Motion blur is a common artifact in image processing, specifically in e-health services, which is caused by the motion of a camera or scene. In linear motion cases, the blur kernel, i.e., the function that simulates the linear motion blur process, depends on the length and direction of blur, called linear motion blur parameters. The estimation of blur parameters is a vital and sensitive stage in the process of reconstructing a sharp version of a motion blurred image, i.e., image deblurring. The estimation of blur parameters can also be used in e-health services. Since medical images may be blurry, this method can be used to estimate the blur parameters and then take an action to enhance the image. In this paper, some methods are proposed for estimating the linear motion blur parameters based on the extraction of features from the given single blurred image. The motion blur direction is estimated using the Radon transform of the spectrum of the blurred image. To estimate the motion blur length, the relation between a blur metric, called NIDCT (Noise-Immune Discrete Cosine Transform-based), and the motion blur length is applied. Experiments performed in this study showed that the NIDCT blur metric and the blur length have a monotonic relation. Indeed, an increase in blur length leads to increase in the blurriness value estimated via the NIDCT blur metric. This relation is applied to estimate the motion blur. The efficiency of the proposed method is demonstrated by performing some quantitative and qualitative experiments.

Effective Blind Image Deblurring Using Matrix-Variable Optimization

Blind image deblurring has been a challenging issue due to the unknown blur and computation problem. Recently, the matrix-variable optimization method successfully demonstrates its potential advantages in computation. This paper proposes an effective matrix-variable optimization method for blind image deblurring. Blur kernel matrix is exactly decomposed by a direct SVD technique. The blur kernel and original image are well estimated by minimizing a matrix-variable optimization problem with blur kernel constraints. A matrix-type alternative iterative algorithm is proposed to solve the matrix-variable optimization problem. Finally, experimental results show that the proposed blind image deblurring method is much superior to the state-of-the-art blind image deblurring algorithms in terms of image quality and computation time.

Layer-Output Guided Complementary Attention Learning for Image Defocus Blur Detection

Defocus blur detection (DBD), which has been widely applied to various fields, aims to detect the out-of-focus or in-focus pixels from a single image. Despite the fact that the deep learning based methods applied to DBD have outperformed the hand-crafted feature based methods, the performance cannot still meet our requirement. In this paper, a novel network is established for DBD. Unlike existing methods which only learn the projection from the in-focus part to the ground-truth, both in-focus and out-of-focus pixels, which are completely and symmetrically complementary, are taken into account. Specifically, two symmetric branches are designed to jointly estimate the probability of focus and defocus pixels, respectively. Due to their complementary constraint, each layer in a branch is affected by an attention obtained from another branch, effectively learning the detailed information which may be ignored in one branch. The feature maps from these two branches are then passed through a unique fusion block to simultaneously get the two-channel output measured by a complementary loss. Additionally, instead of estimating only one binary map from a specific layer, each layer is encouraged to estimate the ground truth to guide the binary map estimation in its linked shallower layer followed by a top-to-bottom combination strategy, gradually exploiting the global and local information. Experimental results on released datasets demonstrate that our proposed method remarkably outperforms state-of-the-art algorithms.

Eliminating unbalanced defocus blur with a binocular linkage network

In this paper, we address the issue of the unbalanced defocus blur problem in stereo vision by a deblurring binocular linkage network. A similarity-enhanced loss function, which regularizes the difference between the output images by disparity warping, is proposed for the left-right sharpness consistency of the outputs. A high match rate is obtained. We test our methods on both synthetic and real data. The experimental results show that our method outperforms the state-of-the-art single and stereo deblurring methods for high accuracy in stereo matching, which is very helpful for long-distance stereo vision measurement.

Image Deblurring Using Multi-Stream Bottom-Top-Bottom Attention Network and Global Information-Based Fusion and Reconstruction Network

Image deblurring has been a challenging ill-posed problem in computer vision. Gaussian blur is a common model for image and signal degradation. The deep learning-based deblurring methods have attracted much attention due to their advantages over the traditional methods relying on hand-designed features. However, the existing deep learning-based deblurring techniques still cannot perform well in restoring the fine details and reconstructing the sharp edges. To address this issue, we have designed an effective end-to-end deep learning-based non-blind image deblurring algorithm. In the proposed method, a multi-stream bottom-top-bottom attention network (MBANet) with the encoder-to-decoder structure is designed to integrate low-level cues and high-level semantic information, which can facilitate extracting image features more effectively and improve the computational efficiency of the network. Moreover, the MBANet adopts a coarse-to-fine multi-scale strategy to process the input images to improve image deblurring performance. Furthermore, the global information-based fusion and reconstruction network is proposed to fuse multi-scale output maps to improve the global spatial information and recurrently refine the output deblurred image. The experiments were done on the public GoPro dataset and the realistic and dynamic scenes (REDS) dataset to evaluate the effectiveness and robustness of the proposed method. The experimental results show that the proposed method generally outperforms some traditional deburring methods and deep learning-based state-of-the-art deblurring methods such as scale-recurrent network (SRN) and denoising prior driven deep neural network (DPDNN) in terms of such quantitative indexes as peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) and human vision.

Efficient learning-based blur removal method based on sparse optimization for image restoration

In imaging systems, image blurs are a major source of degradation. This paper proposes a parameter estimation technique for linear motion blur, defocus blur, and atmospheric turbulence blur, and a nonlinear deconvolution algorithm based on sparse representation. Most blur removal techniques use image priors to estimate the point spread function (PSF); however, many common forms of image priors are unable to exploit local image information fully. In this paper, the proposed method does not require models of image priors. Further, it is capable of estimating the PSF accurately from a single input image. First, a blur feature in the image gradient domain is introduced, which has a positive correlation with the degree of blur. Next, the parameters for each blur type are estimated by a learning-based method using a general regression neural network. Finally, image restoration is performed using a half-quadratic optimization algorithm. Evaluation tests confirmed that the proposed method outperforms other similar methods and is suitable for dealing with motion blur in real-life applications.

Deep Multi-scale Discriminative Networks for Double JPEG Compression Forensics

As JPEG is the most widely used image format, the importance of tampering detection for JPEG images in blind forensics is self-evident. In this area, extracting effective statistical characteristics from a JPEG image for classification remains a challenge. Effective features are designed manually in traditional methods, suggesting that extensive labor-consuming research and derivation is required. In this article, we propose a novel image tampering detection method based on deep multi-scale discriminative networks (MSD-Nets). The multi-scale module is designed to automatically extract multiple features from the discrete cosine transform (DCT) coefficient histograms of the JPEG image. This module can capture the characteristic information in different scale spaces. In addition, a discriminative module is also utilized to improve the detection effect of the networks in those difficult situations when the first compression quality ( QF 1) is higher than the second one ( QF 2). A special network in this module is designed to distinguish the small statistical difference between authentic and tampered regions in these cases. Finally, a probability map can be obtained and the specific tampering area is located using the last classification results. Extensive experiments demonstrate the superiority of our proposed method in both quantitative and qualitative metrics when compared with state-of-the-art approaches.

Local Restricted Convolutional Neural Network for Change Detection in Polarimetric SAR Images

To detect changed areas in multitemporal polarimetric synthetic aperture radar (SAR) images, this paper presents a novel version of convolutional neural network (CNN), which is named local restricted CNN (LRCNN). CNN with only convolutional layers is employed for change detection first, and then LRCNN is formed by imposing a spatial constraint called local restriction on the output layer of CNN. In the training of CNN/LRCNN, the polarimetric property of SAR image is fully used instead of manual labeled pixels. As a preparation, a similarity measure for polarimetric SAR data is proposed, and several layered difference images (LDIs) of polarimetric SAR images are produced. Next, the LDIs are transformed into discriminative enhanced LDIs (DELDIs). CNN/LRCNN is trained to model these DELDIs by a regression pretraining, and then a classification fine-tuning is conducted with some pseudolabeled pixels obtained from DELDIs. Finally, the change detection result showing changed areas is directly generated from the output of the trained CNN/LRCNN. The relation of LRCNN to the traditional way for change detection is also discussed to illustrate our method from an overall point of view. Tested on one simulated data set and two real data sets, the effectiveness of LRCNN is certified and it outperforms various traditional algorithms. In fact, the experimental results demonstrate that the proposed LRCNN for change detection not only recognizes different types of changed/unchanged data, but also ensures noise insensitivity without losing details in changed areas.

Hierarchical Extended Bilateral Motion Estimation-Based Frame Rate Upconversion Using Learning-Based Linear Mapping

We present a novel and effective learning-based frame rate upconversion (FRUC) scheme, using linear mapping. The proposed learning-based FRUC scheme consists of: 1) a new hierarchical extended bilateral motion estimation (HEBME) method; 2) a light-weight motion deblur (LWMD) method; and 3) a synthesis-based motion-compensated frame interpolation (S-MCFI) method. First, the HEBME method considerably enhances the accuracy of the motion estimation (ME), which can lead to a significant improvement of the FRUC performance. The proposed HEBME method consists of two ME pyramids with a three-layered hierarchy, where the motion vectors (MVs) are searched in a coarse-to-fine manner via each pyramid. The found MVs are further refined in an enhanced resolution of four times by jointly combining the MVs from the two pyramids. The HEBME method employs a new elaborate matching criterion for precise ME which effectively combines a bilateral absolute difference, an edge variance, pixel variances, and an MV difference among two consecutive blocks and its neighboring blocks. Second, the LWMD method uses the MVs found by the HEBME method and removes the small motion blurs in original frames via transformations by linear mapping. Third, the S-MCFI method finally generates interpolated frames by applying linear mapping kernels for the deblurred original frames. In consequence, our FRUC scheme is capable of precisely generating interpolated frames based on the HEBME for accurate ME, the S-MCFI for elaborate frame interpolation, and the LWMD for contrast enhancement. The experimental results show that our FRUC significantly outperforms the state-of-the-art non-deep learning-based schemes with an average of 1.42 dB higher in the peak signal-to-noise-ratio and shows comparable performance with the state-of-the-art deep learning-based scheme.

Learning Converged Propagations with Deep Prior Ensemble for Image Enhancement

Enhancing visual qualities of images plays very important roles in various vision and learning applications. In the past few years, both knowledge-driven maximum a posterior (MAP) with prior modelings and fully data-dependent convolutional neural network (CNN) techniques have been investigated to address specific enhancement tasks. In this paper, by exploiting the advantages of these two types of mechanisms within a complementary propagation perspective, we propose a unified framework, named deep prior ensemble (DPE), for solving various image enhancement tasks. Specifically, we first establish the basic propagation scheme based on the fundamental image modeling cues and then introduce residual CNNs to help predicting the propagation direction at each stage. By designing prior projections to perform feedback control, we theoretically prove that even with experience-inspired CNNs, DPE is definitely converged and the output will always satisfy our fundamental task constraints. The main advantage against conventional optimization-based MAP approaches is that our descent directions are learned from collected training data, thus are much more robust to unwanted local minimums. While, compared with existing CNN type networks, which are often designed in heuristic manners without theoretical guarantees, DPE is able to gain advantages from rich task cues investigated on the bases of domain knowledges. Therefore, DPE actually provides a generic ensemble methodology to integrate both knowledge and data-based cues for different image enhancement tasks. More importantly, our theoretical investigations verify that the feedforward propagations of DPE are properly controlled toward our desired solution. Experimental results demonstrate that the proposed DPE outperforms state-of-the-arts on a variety of image enhancement tasks in terms of both quantitative measure and visual perception quality.

Blind deblurring of natural stochastic textures using an anisotropic fractal model and phase retrieval algorithm

The challenging inverse problem of blind deblurring has been investigated thoroughly for natural images. Existing algorithms exploit edge-type structures, or similarity to smaller patches within the image, to estimate the correct blurring kernel. However, these methods do not perform well enough on natural stochastic textures (NST), which are mostly random and in general are not characterized by distinct edges and contours. In NST even small kernels cause severe degradation to images. Restoration poses therefore an outstanding challenge. In this work, we refine an existing method by implementing an anisotropic fractal model to estimate the blur kernel's power spectral density. The final kernel is then estimated via an adaptation of a phase retrieval algorithm, originally proposed for sparse signals. We further incorporate additional constraints that are specific to blur filters, to yield even better results. The latter are compared with results obtained by recently published blind deblurring methods.

Training Very Deep CNNs for General Non-Blind Deconvolution

Non-blind image deconvolution is an ill-posed problem. The presence of noise and band-limited blur kernels makes the solution of this problem non-unique. Existing deconvolution techniques produce a residual between the sharp image and the estimation that is highly correlated with the sharp image, the kernel, and the noise. In most cases, different restoration models must be constructed for different blur kernels and different levels of noise, resulting in low computational efficiency or highly redundant model parameters. Here we aim to develop a single model that handles different types of kernels and different levels of noise: general non-blind deconvolution. Specifically, we propose a very deep convolutional neural network that predicts the residual between a pre-deconvolved image and the sharp image rather than the sharp image. The residual learning strategy makes it easier to train a single model for different kernels and different levels of noise, encouraging high effectiveness and efficiency. Quantitative evaluations demonstrate the practical applicability of the proposed model for different blur kernels. The model also shows state-of-the-art performance on synthesized blurry images.

Pairwise Operator Learning for Patch-Based Single-Image Super-Resolution

Motivated by the fact that image patches could be inherently represented by matrices, single-image super-resolution is treated as a problem of learning regression operators in a matrix space in this paper. The regression operators that map low-resolution image patches to high-resolution image patches are generally defined by the left and right multiplication operators. The pairwise operators are, respectively, used to extract the raw and column information of low-resolution image patches for recovering high-resolution estimations. The patch-based regression algorithm possesses three favorable properties. First, the proposed super-resolution algorithm is efficient during both training and testing, because image patches are treated as matrices. Second, the data storage requirement of the optimal pairwise operator is far less than most popular single-image super-resolution algorithms, because only two small sized matrices need to be stored. Last, the super-resolution performance is competitive with most popular single-image super-resolution algorithms, because both raw and column information of image patches is considered. Experimental results show the efficiency and effectiveness of the proposed patch-based single-image super-resolution algorithm.