Improved TV for blind deblurring with split Bregman iteration

In this paper, an improved TV (total variation) with split Bregman iteration is proposed to improve the result of blind image deblurring. According to the property of the higher partial derivatives in the local image, adaptive values of regularisation parameters in TV algorithm are calculated in order to suppress the ringing artefacts effectively. To alleviate the difficulty of L₁ norm problem, the split Bregman iteration is also adopted. Comparisons are made between the proposed and traditional methods, and experimental results prove that the proposed algorithm is not only robust to different size of blur kernels, but also can suppress the ringing artefacts to a great extent in the latent image, both visually and theoretically.     

基于颜色空间变换和CNN的自适应去模糊方法

目的研究数字图像中的去模糊问题,从受损的模糊图像中恢复出清晰图像。方法针对现有图像去模糊算法无法保留图像高频信息及容易产生振铃效应等问题,提出一种基于Y通道反卷积和卷积神经网络的两阶段自适应去模糊算法(SDYCNN)。在第1阶段,将数字图像转换至YUV颜色空间,根据图像无参考质量评价分数与模糊核尺寸之间的对应关系,在Y通道内自适应确定模糊核尺寸并进行反卷积增强;第2阶段将第1阶段中的反卷积增强作为预处理方式,通过4层卷积神经网络建立反卷积增强后的图像与清晰图像之间的映射关系,实现图像去模糊。结果轻微模糊图像在第1阶段便能够得到较好的去模糊效果,严重模糊图像经过第1阶段的反卷积增强,也有助于神经网络中特征的快速提取。结论实验结果表明,该算法不仅对于模糊图像具有良好的恢复效果,运算效率也有显著提升。     

基于强度和梯度先验的L0正则化模糊QR码识别

目的研究因机械抖动,拍摄器材与图像存在一定距离或相对运动而产生运动模糊、散焦模糊等情况下的模糊QR码图像识别。方法采用基于强度和梯度先验的L_0正则化方法对模糊QR图像进行去模糊。优化模糊核尺寸的人为预估问题,提高程序效率。对1至15类常用QR码图像进行模糊仿真,再通过盲提取获得模糊核,用峰值信噪比PSNR值衡量该方法在QR码图像去模糊的复原精度。结果PSNR值随着QR码图像复杂度的增加而相对减少,但因QR码存在一定的容错率,在PSNR值为13以上且噪声、振铃小的情况下就能够被识别。文中算法相较于其他算法在型号较高的模糊QR码恢复方面识别率更高。结论基于强度和梯度先验的L0正则化方法对模糊QR码的恢复效果显著,且不是只针对某一类模糊QR码图像,对于多种类型的模糊QR码图像恢复都能有很好的效果。     

Image deblurring using empirical Wiener filter in the curvelet domain and joint non-local means filter in the spatial domain

In this paper, an efficient image deblurring algorithm is proposed. This algorithm restores the blurred image by incorporating a curvelet-based empirical Wiener filter with a spatial-based joint non-local means filter. Curvelets provide a multidirectional and multiscale decomposition that has been mathematically shown to represent distributed discontinuities such as edges better than traditional wavelets. Our method restores the image in the frequency domain to obtain a noisy result with minimal loss of image components, followed by an empirical Wiener filter in the curvelet domain to attenuate the leaked noise. Although the curvelet-based methods are efficient in edge-preserving image denoising, they are prone to producing edge ringing which relates to the structure of the underlying curvelet. In order to reduce the ringing, we develop an efficient joint non-local means filter by using the curvelet deblurring result. This filter could suppress the leaked noise while preserving image details. We compare our deblurring algorithm with a few competitive deblurring techniques in terms of improvement in signal-to-noise-ratio (ISNR) and visual quality.     

基于深度强化学习的图像去模糊方法

目的为了有效地去除多种图像模糊,提高图像质量,提出基于深度强化学习的图像去模糊方法。方法选用GoPro与DIV2K这2个数据集进行实验,以峰值信噪比(PSNR)和结构相似性(SSIM)为客观评价指标。通过卷积神经网络获得模糊图像的高维特征,利用深度强化学习结合多种CNN去模糊工具建立去模糊框架,将峰值信噪比(PSNR)作为训练奖励评价函数,来选择最优修复策略,逐步对模糊图像进行修复。结果通过训练与测试,与现有的主流算法相比,文中方法有着更好的主观视觉效果,且PSNR值与SSIM值都有更好的表现。结论实验结果表明,文中方法能有效地解决图像的高斯模糊和运动模糊等问题,并取得了良好的视觉效果,在图像去模糊领域具有一定的参考价值。     

基于超拉普拉斯先验与核谱特性噪声图像盲去模糊

现有大部分盲图像去模糊方法对噪声敏感,即使少量的噪声可大大降低恢复图像的质量．考虑到模糊图像中同时隐含有清晰图像信息和模糊核信息,我们同时利用卷积核谱特性先验和清晰图像梯度域超拉普拉斯先验联合建立含噪图像盲去模糊模型,较单独使用卷积核先验与清晰图像先验建模更合理,也能获得更精确的估计图像．本文借助于Hessian矩阵,利用模糊图像及卷积核联合生成先验子,而非单独的估计图像先验子,建立优化模型．求解模型时,通过迭代策略交替细化模糊核和清晰图像．在清晰图像恢复阶段,因存在超拉普拉斯先验项,提出用变量分离法计算清晰图像．清晰图像采用快速傅里叶变换及封闭阈值公式求解,以提高优化速度．实验结果表明：与其他方法相比,本文方法能获得更鲁棒的模糊核和更精确的清晰图像,且收敛速度更快．     

基于散斑图像梯度域的数字图像相关盲去模糊方法EI北大核心CSCD

针对复杂测量环境或高动态测量过程中出现的运动模糊问题,提出了一种灰度稀疏先验与参考图像梯度域先验相结合的散斑图像盲去模糊方法。该方法以灰度直方图峰值的L 0范数与参考图像梯度域分布建立优化函数正则项,使用二次分裂方法估计清晰图像,再以交替迭代的方式进行卷积核细化。在模糊核估计完成后,使用Richardson-Lucy非盲去卷积方法完成散斑图像的复原。实验结果证明:所提出的散斑图像盲去模糊方法与针对自然图像与文本图像的经典方法相比,获得了更优的图像去模糊效果,并提高了数字图像相关测量精度与鲁棒性。     

Iterative statistical approach to blind image deconvolution

Image deblurring has long been modeled as a deconvolution problem. In the literature, the point-spread function (PSF) is often assumed to be known exactly. However, in practical situations such as image acquisition in cameras, we may have incomplete knowledge of the PSF. This deblurring problem is referred to as blind deconvolution. We employ a statistical point of view of the data and use a modified maximum a posteriori approach to identify the most probable object and blur given the observed image. To facilitate computation we use an iterative method, which is an extension of the traditional expectation-maximization method, instead of direct optimization. We derive separate formulas for the updates of the estimates in each iteration to enhance the deconvolution results, which are based on the specific nature of our a priori knowledge available about the object and the blur.     

A novel blind super‐resolution technique based on the improved Poisson maximum a posteriori algorithm

Image restoration has received considerable attention. In many practical situations, unfortunately, the blur is often unknown, and little information is available about the true image. Therefore, the true image is identified directly from the corrupted image by using partial or no information about the blurring process and the true image. In addition, noise will be amplified to induce severely ringing artifacts in the process of restoration. This article proposes a novel technique for the blind super‐resolution, whose mechanism alternates between de‐convolution of the image and the point spread function based on the improved Poisson maximum a posteriori super‐resolution algorithm. This improved Poisson MAP super‐resolution algorithm incorporates the functional form of a Wiener filter into the Poisson MAP algorithm operating on the edge image further to reduce noise effects and speed restoration. Compared with that based on the Poisson MAP, the novel blind super‐resolution technique presents experimental results from 1‐D signals and 2‐D images corrupted by Gaussian point spread functions and additive noise with significant improvements in quality. © 2003 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 12, 239–246, 2002; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ima.10032     

Defocus deblurring for catadioptric omnidirectional imaging based on spatially invariant point spread function

The problem of catadioptric omnidirectional imaging defocus blur, which is caused by lens aperture and mirror curvature, becomes more severe when high resolution sensors and large apertures are applied. In order to overcome this problem, a novel method based on computational photography is proposed. Firstly, the defocus blur of catadioptric omnidirectional imaging is analyzed to calculate the point spread function for different scene points. Then, the defocus blur kernel of omnidirectional image is confirmed to be spatially invariant when rotating the focus ring of camera lens during an image’s integration time. Lastly, the deconvolution algorithm using prior sparse derivatives is applied to obtain all-focused/sharp omnidirectional images. Experimental results demonstrate that the proposed method is effective for omnidirectional image deblurring and can be applied to most existing catadioptric omnidirectional imaging systems.     

贝叶斯推理模型耦合非平稳边缘保持先验的图像模糊消除

目的针对现行图像去模糊消除机制忽略了图像空间结构特征,降低了模糊消除效果,且算法稳定性不佳,无法克服解模糊等的不足,提出了贝叶斯模型耦合非平稳先验的图像去模糊机制。方法基于二阶统计量方法,定义模糊函数;引入滤波因子和超参数,构造非平稳边缘保持先验模型;基于贝叶斯推理,引入雅克比矩阵设计了超参数动态更新机制;用耦合先验模型与贝叶斯模型完成图像复原。在仿真平台上测试了算法的性能。结果与其他几种机制相比,提出的算法机制去模糊质量更好,局部放大后纹理细节仍然清晰,并且去模糊前后图像的结构相似度更高。结论提出的算法具有较佳的图像去模糊效果,重构质量理想。     

变量分裂机制耦合非循环模糊模型的图像复原

目的针对当前非二次正则化图像复原算法在图像边界的周期拓展难度较大,导致复原图像的边界存在振铃效应,算法的通用性削弱,且时耗严重的问题,提出变量分裂机制耦合非循环卷积模型的抗失真图像快速复原算法。方法引入掩模算子,设计非循环模糊模型,显著消除复原图像的振铃效应;改进数据保真项,构造了最小成本化函数;定义二次辅助变量,嵌入变量分裂策略,设计基于变量分裂的乘数交替方向算法,对非循环模糊模型中的循环与掩模矩阵进行解耦,降低算法复杂度;构造增广拉格朗日函数,耦合交替最小机制与成本函数,以单步封闭式更新估算图像,快速完成图像重构。结果仿真结果显示:与当前复原算法相比,提出算法的失真度最小,且收敛速度更快。结论提出算法能够快速复原多种类型的退化图像,有效消除了复原图像的振铃效应。     

基于生成对抗网络的QR码去模糊算法

目的 针对包装产品上QR码在采集过程中的运动模糊、失焦模糊,长期磨损形成的自模糊和环境中的噪声等因素,导致QR码无法识别的问题,提出一种基于生成对抗网络的QR码去模糊算法。方法 采用深度学习模型生成对抗网络对模糊核和环境噪声具有的强大拟合和估计能力,提取模糊QR码图像与真实图像的深层特征和差距,并通过生成器与判别器不断迭代对抗,使生成器具有由输入的模糊QR码产生与之对应的去模糊QR码图像的能力。结果 生成器能较好地对模糊核和环境噪声进行估计,而且能够实现对数据集内多种不同模糊程度QR码的去模糊,去模糊QR码图像效果较好,处理时间快,识别率较高。结论 采用基于生成对抗网络的QR码去模糊算法能够广泛应用于包装产品外壳上QR码的预处理过程,泛化能力较好,能有效提高扫描识别率。     

Feasibility Study for Improving the Image Characteristics in Digital Tomosynthesis (DTS) Using a Compressed-Sensing (Cs)-Based Pre-Deblurring Scheme

Digital tomosynthesis (DTS) has been widely used in both industrial nondestructive testing and medical x-ray imaging as a popular multiplanar imaging modality. However, although it provides some of the tomographic benefits of computed tomography (CT) at reduced dose and imaging time, the image characteristics are relatively poor due to blur artifacts originated from incomplete data sampling for a limited angular range and also aspects inherent to imaging system, including finite focal spot size of the x-ray source, detector resolution, etc. In this work, in order to overcome these difficulties, we propose an intuitive method in which a compressed-sensing (CS)-based deblurring scheme is applied to the projection images before common DTS reconstruction. We implemented the proposed deblurring algorithm and performed a systematic experiment to demonstrate its viability for improving the image characteristics in DTS. According to our results, the proposed method appears to be effective for the blurring problems in DTS and seems to be promising to our ongoing application to x-ray nondestructive testing.     

Parallel-distributed blind deconvolution based on a self-organizing neural network

A parallel-distributed blind deconvolution method based on a self-organizing neural network is introduced. A large degraded image is segmented into smaller subpatterns. Each subpattern can be used to get a blur function. Moreover, we propose a two-step unsupervised learning method in the self-organizing neural network. The two-step learning method includes parallel learning and series learning operations. The series learning operation is similar to a typical learning operation in the self-organizing neural network. The parallel learning operation is used as a positive perturbation to let the learning operation leave a local minimum. Several improved blur functions can be estimated from the different subpatterns, and the optimized blur function is evolved by use of a genetic algorithm. As the blur function is estimated, the source image of the large degraded image can be easily restored by use of a Wiener-type filter or other deconvolution methods. Computer simulations show that the proposed parallel-distributed blind deconvolution method gives good reconstruction and that the two-step learning method in the self-organizing neural network can promote learning. Since the main computational cost is dependent on the size of the subpattern, the proposed method is effective for the restoration of the large image.     

Estimation of perceived image blur using edge features

Sharpness (or its complement, perceived blur or unsharpness) is an important attribute of image quality, and the spread of the physical blurring kernel is the predominant parameter determining that attribute. In this article we present an algorithm to estimate an objective measure for sharpness, called the blur index. The algorithm first estimates the physical parameter of blur spread from the blurred image and subsequently uses that estimate to compute the blur index. A global estimate of blur spread for the entire image is obtained by the weighted averaging of the local estimates of blur spread at prominent edge locations in the image. These local estimates at edges are obtained by nonlinearly combining local derivatives. The edge prominence is based on the edge height and the edge-contour length. The blur index is computed from the estimated blur spread by taking the sensitivity of the visual system to changes in the blur spread into account. The results of a psychophysical experiment in which subjects judged the unsharpness of natural images are also reported. By correlating the estimates of the blur index, as obtained from the algorithm, with the results obtained in the psychophysical experiment, we show that the blur index correlates well with the perceived unsharpness, and hence can be considered a psychometric measure of sharpness. © 1996 John Wiley & Sons, Inc.     

用多重菲涅耳衍射变换实现盲信息隐藏

提出了一种新的基于多重菲涅耳衍射变换和像素替代的盲信息隐藏算法.需隐藏的图像经多重菲涅耳衍射变换为一密文复矩阵,然后将其实部和虚部分别嵌入到原始宿主图像中,同时将此已嵌入信息的原像素值用其近邻的未嵌入信息的像素均值来替代,从而实现盲信息隐藏.数值仿真计算结果表明:该隐藏算法对JPEG有损压缩、图像剪切,噪声污染、重采样攻击和亮度、对比度、直方图、灰度曲线调整等具有一定的抵抗能力;由于采用一系列加密密钥(光波长、透镜焦距,多个衍射距离等),只有当所有密钥都正确时,才能解密恢复所隐藏的信息,所以该算法具有较强的鲁棒性和很高的安全性.     

Robust and computationally efficient superresolution algorithm

Superresolution is the process of combining information from multiple subpixel-shifted low-resolution images to form a high-resolution image. It works quite well under ideal conditions but deteriorates rapidly with inaccuracies in motion estimates. We model the original high-resolution image as a Markov random field (MRF) with a discontinuity adaptive regularizer. Given the low-resolution observations, an estimate of the superresolved image is obtained by using the iterated conditional modes (ICM) algorithm, which maximizes the local posterior conditional probability sequentially. The proposed method not only preserves edges but also lends robustness to errors in the estimates of motion and blur parameters. We derive theoretically the neighborhood structure for the posterior distribution in the presence of warping, blurring, and downsampling operations and use this to effectively reduce the overall computations. Results are given on synthetic as well as real data to validate our method.     

基于Z变换匀速旋转运动模糊图像的快速恢复

从离散直观角度分析了由匀速旋转运动引起图像模糊的机理,通过Z变换方法,建立了由匀速旋转运动引起图像模糊的退化模型和恢复模型,并给出了快速算法。将运动模型扩展到任意形式的匀速运动,并给出了数学模型。通过计算机仿真实验,验证了算法的正确性、有效性和快速性。本文推导的是理想模型,没有考虑噪声的影响。     

Restoration of images from the scanning-tunneling microscope

During the acquisition of an image from any probe microscope instrument, various noise sources cause distortion in the observed image. It is often the case that impulsive disturbances cause bright groups of pixels to replace the actual image data in these locations. Furthermore, the images from a probe microscope show some amount of blurring caused both by the instrument function and the material properties. In almost all image-processing applications it is important to remove any impulsive distortion that may be present before deblurring can be attempted. We give a technique for detecting these impulses and reconstructing the image. This technique is superior to the standard global application of median filters for the case considered. The reconstruction is limited only to the affected regions and therefore results in a much sharper and more meaningful image. With the assumption of Gaussian blur it is then possible to propose several different deblurring methodologies. We present a novel Wiener-filter deblurring implementation and compare it to both maximum-entropy and Richardson-Lucy deblurring.