一种非凸核范数最小化一般模型及其在图像去噪中的应用

聚焦于非凸的低秩逼近模型,提出了一类定义在矩阵奇异值上的非凸函数g,实际上很多著名的非凸函数都满足g函数的条件。将g函数引入带权核范数最小化模型得到更一般的模型,可以很好地解决模型中的权重选择问题。将该模型应用于图像去噪领域,并针对该模型给出收敛的求解算法。仿真实验表明, 相对其他先进的算法所提方法更具优势。     

Weighted Nuclear Norm Minimization and Its Applications to Low Level Vision

As a convex relaxation of the rank minimization model, the nuclear norm minimization (NNM) problem has been attracting significant research interest in recent years. The standard NNM regularizes each singular value equally, composing an easily calculated convex norm. However, this restricts its capability and flexibility in dealing with many practical problems, where the singular values have clear physical meanings and should be treated differently. In this paper we study the weighted nuclear norm minimization (WNNM) problem, which adaptively assigns weights on different singular values. As the key step of solving general WNNM models, the theoretical properties of the weighted nuclear norm proximal (WNNP) operator are investigated. Albeit nonconvex, we prove that WNNP is equivalent to a standard quadratic programming problem with linear constrains, which facilitates solving the original problem with off-the-shelf convex optimization solvers. In particular, when the weights are sorted in a non-descending order, its optimal solution can be easily obtained in closed-form. With WNNP, the solving strategies for multiple extensions of WNNM, including robust PCA and matrix completion, can be readily constructed under the alternating direction method of multipliers paradigm. Furthermore, inspired by the reweighted sparse coding scheme, we present an automatic weight setting method, which greatly facilitates the practical implementation of WNNM. The proposed WNNM methods achieve state-of-the-art performance in typical low level vision tasks, including image denoising, background subtraction and image inpainting.     

Nonlocal image denoising via adaptive tensor nuclear norm minimization

Nonlocal self-similarity shows great potential in image denoising. Therefore, the denoising performance can be attained by accurately exploiting the nonlocal prior. In this paper, we model nonlocal similar patches through the multi-linear approach and then propose two tensor-based methods for image denoising. Our methods are based on the study of low-rank tensor estimation (LRTE). By exploiting low-rank prior in the tensor presentation of similar patches, we devise two new adaptive tensor nuclear norms (i.e., ATNN-1 and ATNN-2) for the LRTE problem. Among them, ATNN-1 relaxes the general tensor N-rank in a weighting scheme, while ATNN-2 is defined based on a novel tensor singular-value decomposition (t-SVD). Both ATNN-1 and ATNN-2 construct the stronger spatial relationship between patches than the matrix nuclear norm. Regularized by ATNN-1 and ATNN-2 respectively, the derived two LRTE algorithms are implemented through the adaptive singular-value thresholding with global optimal guarantee. Then, we embed the two algorithms into a residual-based iterative framework to perform nonlocal image denoising. Experiments validate the rationality of our tensor low-rank assumption, and the denoising results demonstrate that our proposed two methods are exceeding the state-of-the-art methods, both visually and quantitatively.

     

基于非凸低秩约束的图像修复方法

受传输干扰或存储不当等因素的影响,现实应用中获取的某些图像通常会存在像素缺失现象,这给图像的后续分析与处理带来了一定影响.解决该问题的常用方法是对图像进行低秩修复.利用低秩特性进行修复的方法大多以秩函数建模,由于矩阵秩函数是非凸离散的,该模型的求解是一个NP难问题,所以通常利用核范数对矩阵的秩进行凸松弛.但是,基于核范...     

基于带权核范数最小化和混合高斯模型的去噪模型

非局部自相似性（NSS）先验在图像恢复中发挥重要作用,如何充分利用这一先验提高图像恢复性能仍值得深入研究,提出一种基于带权核范数最小化和混合高斯模型的去噪模型。首先,采用混合高斯模型（GMM）对无噪声的自然图像非局部自相似图像块进行训练,再用训练好的混合高斯模型指导退化的图像产生非局部自相似图像块组;然后,结合带权的核范数最小化技术实现图像的去噪,并对模型的保真项进行一般性扩展,给出收敛的求解算法。仿真实验表明,所提方法与基于3D滤波的块匹配（BM3D）算法、同时稀疏编码学习（LSSC）算法和带权的核范数最小化（WNNM）模型相比,峰值信噪比（PSNR）提高0.11~0.49 dB。     

截断核范数和全变差正则化高光谱图像复原

目的 高光谱图像距具有较高的光谱分辨率,从而具备区分诊断性光谱特征地物的能力,但高光谱数据经常会受到如环境、设备等各种因素的干扰,导致数据污染,严重影响高光谱数据在应用中的精度和可信度。方法 根据高光谱图像光谱维度特征值大小与所包含信息的关系,利用截断核范数最小化方法表示光谱低秩先验,从而有效抑制稀疏噪声;再利用高光谱图像的空间稀疏先验建立正则化模型,达到去除高密度噪声的目的;最终,结合上述两种模型的优势,构建截断核范数全变差正则化模型去除高斯噪声、稀疏噪声及其他混合噪声等。结果 将本文与其他三种近期发表的主流去噪方法进行对比,模型平均峰信噪比提高3.20 dB,平均结构相似数值指标提高0.22,并可以应用到包含各种噪声、不同尺寸的图像,其模型平均峰信噪比提高1.33 dB。结论 本文方法在光谱低秩中更加准确地表示了观测数据的先验特征,利用高光谱遥感数据的空间和低秩先验信息,能够对含有高密度噪声以及稀疏异常值的图像进行复原。     

总变差正则化断层图像重建的解耦Bregman迭代算法

在断层重建的很多工程应用中,由于低剂量以及成像硬件等原因,经常需要在测量数据不充分的情况下去重建图像。基于图像分段光滑的假设,提出采用误差的加权范数作为数据保真项,TV（total variation）作为正则项的断层图像重建模型。该模型求解时,首先通过引入代理函数将原问题解耦为残差的加权范数最小化和加权范数TV去噪这两个子问题;然后采用了Chambolle的对偶空间正交投影法的框架对加权范数TV去噪问题进行求解,避免了由于TV项在不可导处所带来的计算不稳定;最后,为了提高收敛速度并且避免由正则化参数选取所引起的数值不稳定,引入Bregman方法,给出该模型的快速迭代算法。在扇形束少角度欠采样的条件下,对理想情况和高斯噪声情况下进行仿真测试,并同多种算法进行了比较。实验结果表明,该算法重建效果好,收敛速度快。     

联合矩阵F范数的低秩图像去噪

摘 要：目的：低秩矩阵恢复是通过最小化矩阵核范数来获得低秩解,然而待恢复低秩矩阵相关性低的要求往往会导致求解不稳定的情况。方法：针对该问题,研究一种基于变量分裂的低秩图像恢复去噪算法,引入待恢复矩阵的Frobenius范数作为新正则项,与原有低秩矩阵的核范数组成联合正则化项,对问题进行凸松弛后,采用变量分裂的增广拉格朗日乘子法求解。结果：为考察方法的稳定性和去噪能力,选取了不同参数类型的加噪图像进行仿真,并结合恢复时间、信噪比、差错率等评价标准与现有低秩矩阵恢复算法进行对比。结论：实验结果表明增加Frobenius范数的低秩矩阵恢复模型在保持原有低秩稀疏恢复的前提下,具有良好的去噪性能,对相关性强的低秩图像恢复结果稳定性好,获得了更高的信噪比。     

图像差与加权核范数最小化的压缩图像融合

现有的图像融合算法存在非线性操作产生的噪声干扰和空间复杂度高等问题,使得融合图像易失真和丢失信息。一些学者提出的压缩感知图像融合算法能有效改善这一问题,但大多忽略了图像矩阵的低秩性,往往会降低融合质量。由此,将压缩感知融合技术与低秩矩阵逼近方法相结合,提出基于信息论图像差与自适应加权核范数最小化的图像融合算法。该算法由3个阶段组成。首先,将2幅源图像通过小波稀疏基稀疏化,并利用结构随机矩阵压缩采样,得到测量输出矩阵。然后,将测量输出矩阵进行分块,再利用图像差融合算法得到融合后的测量输出矩阵块。最后,利用自适应加权核范数最小化优化得到的块权重,通过正交匹配追踪法重建融合图像。实验结果表明了该算法的有效性和普适性,并且在多种评价指标上优于其他融合算法。     

加权核范数降噪算法在扩散加权图像中的应用

目的 扩散加权成像技术是一种能够检测活体组织内水分子扩散运动的无创方法,其对数据的准确度要求较高且对噪声较为敏感。扩散加权图像的自相似性程度高,纹理细节较多且纹理和结构具有重复出现的特性。而获取图像的过程中受到不可避免的噪声干扰会破坏图像的数据准确度,因此对扩散加权图像进行降噪是十分必要的。方法 根据扩散加权图像的特点,提出将加权核范数降噪算法应用于扩散加权图像的降噪。加权核范数降噪算法由于能够利用图像的自相似性,通过对图像中的相似块进行处理从而实现对图像的降噪,该算法能够保存图像中大量的纹理细节信息。结果 通过模拟数据实验和真实数据实验,将加权核范数降噪算法与传统的扩散加权图像降噪算法如各向异性算法进行比较,结果表明,加权核范数降噪算法相较于其他算法得到的峰值信噪比至少高出20 dB,结构相似性值也至少高出其他算法0.20.5,再将降噪后的图像进行神经纤维跟踪处理,得到的神经纤维平均长度较其他算法至少要长0.20.8且纤维更为平滑。结论 加权核范数降噪算法不仅能够更好地减少扩散加权图像中的噪声,同时也能够最大限度地保存扩散加权图像的纹理细节,降噪效果理想,提高了数据的准确度及有效性。     

基于频率加权张量核范数的高光谱图像复原

目的 高光谱图像复原是高光谱领域中一个重要的预处理步骤,能够有效去除成像条件所带来的不利影响,提升后续处理任务的精度。张量核范数被广泛应用于高光谱复原问题中,得到了较好的结果。然而,在张量核范数的定义中,它对张量所有奇异值使用相同的阈值进行收缩,未充分考虑高光谱的物理意义,得到了次优的结果。为了提升高光谱图像复原的精度,本文提出了基于频率加权张量核范数的高光谱复原算法。方法 在张量的频率域内,对清晰的高光谱图像添加噪声,图像信息在低频部分变化较小,而在高频部分变化巨大。基于这样的物理意义,定义了一种频率加权张量核范数来逼近张量秩函数,提出了频率域权重的自适应确定方法,让其能减少对低频部分的收缩,同时加大高频部分惩罚。然后将其应用于高光谱图像复原和去噪问题中,并基于交替方向乘子法设计了相应最小化问题的快速求解算法。结果 在4个高光谱数据集上与相关方法进行对比仿真实验,高采样率条件下在Washington DC Mall数据集上,相比性能第2的模型,本文模型复原结果的PSNR （peak signal-to-noise ratio）提升了1.76 dB;在Stuff数据集上,PSNR值提升了2.91 dB。高噪声条件下,在Pavia数据集上相比性能第2的模型,本文模型去噪结果的PSNR提升了8.61 dB;在Indian数据集上,PSNR值提升了10.77 dB。结论 本文模型可以更好地探索高光谱图像的低秩特性,使复原的图像在保持主体信息的同时,复原出更多图像纹理细节。     

Image denoising with norm weighted fusion estimators

In recent era, the weighted matrix rank minimization is used to reduce image noise, promisingly. However, low-rank weighted conditions may cause oversmoothing or oversharpening of the denoised image. This demands a clever engineering algorithm. Particularly, to remove heavy noise in image is always a challenging task, specially, when there is need to preserve the fine edge structures. To attain a reliable estimate of heavy noise image, a norm weighted fusion estimators method is proposed in wavelet domain. This holds the significant geometric structure of the given noisy image during the denoising process. Proposed method is applied on standard benchmark images, and simulation results outperform the most popular rivals of noise reduction approaches, such as BM3D, EPLL, LSSC, NCSR, SAIST, and WNNM in terms of the quality measurement metric PSNR (dB) and structural analysis SSIM indices.     

基于准范数与零范数的矩阵恢复

针对高维度矩阵的低秩恢复问题中核范数与l1范数过惩罚导致的结果偏差,提出一种矩阵恢复方法。使用准范数代替低秩恢复问题中常用的核范数约束,使用零范数代替l1范数约束。对于准范数的求解问题,采用与准范数等价的Frobenius/核混合范数进行替代,提出基于交替近似的线性最小化方法对目标函数进行求解。在合成数据与真实数据上的实验结果表明,该算法在主观视觉效果与客观数值比较上都能获得更好的结果。     

加权低秩矩阵恢复的混合噪声图像去噪

传统的基于低秩矩阵恢复的图像去噪算法只对低秩部分进行约束,当高斯噪声过大时,会导致去噪不充分或细节严重丢失。针对此问题,提出了一种新的鲁棒的图像去噪模型。该模型在原有的低秩矩阵核范数约束的基础上引入高斯噪声约束项,此外为了提高低秩矩阵的低秩性和稀疏矩阵的稀疏性,引入了加权的方法。为了考察方法的去噪能力,选取了不同参数类型的混合噪声图像进行仿真,并结合峰值信噪比、结构相似度评价标准与传统的基于低秩矩阵恢复的图像去噪算法进行对比。实验结果表明,加权低秩矩阵恢复的混合噪声图像去噪算法能增加低秩矩阵的低秩性和稀疏矩阵的稀疏性,在保证去噪效果的同时,保留了图像的细节信息,具有更佳的视觉效果,同时,客观评价指标均有所提高。     

基于稀疏优化lp正则化的光滑化拟牛顿算法

压缩感知被广泛应用于信号恢复和图像重构与去噪,重构算法是压缩感知的关键部分之一。当采样率很低时,重建原始信号是个困难的问题。对此,现有算法普遍表现不佳。采用[p(0相似文献   

Hammerstein system identification using nuclear norm minimization

This paper presents a new method for the identification of Hammerstein systems. The parameter estimation problem is formulated as a rank minimization problem by constraining a finite dimensional time dependency between signals. Due to the unknown intermediate signal, the rank minimization problem cannot be solved directly. Thus, the rank minimization problem is reformulated as an intermediate signal construction problem. The main assumption used in this paper is that static nonlinearity is monotonically non-decreasing in order to guarantee a unique combination of a static nonlinear block and a Finite Impulse Response (FIR) linear block. The rank minimization is then relaxed to a convex optimization problem using a nuclear norm. The main contribution of this paper is that the proposed method extends the rank minimization approach to Hammerstein system identification, and does not need a bilinear parametrization and singular value decomposition (SVD), which are commonly used in two-step approaches for Hammerstein system identification.     

时空RPCA在复杂场景下的运动目标检测

在复杂动态背景下,鲁棒主成分分析模型(RPCA)容易将背景中动态背景误判为前景运动目标,导致运动目标检测精度不高。为解决该问题,提出一种基于非凸加权核范数的时空低秩RPCA算法。使用非凸加权核范数替代传统的核范数进行低秩约束,在观测矩阵上通过拉普拉斯特征映射得到时空图拉普拉斯矩阵,将得到的时空图拉普拉斯矩阵嵌入低秩背景矩阵以保持背景对噪声和离群值的鲁棒性。实验结果表明,所提模型在复杂场景中能较准确检测出运动目标。     

Non‐Local Image Inpainting Using Low‐Rank Matrix Completion

In this paper, we propose a highly accurate inpainting algorithm which reconstructs an image from a fraction of its pixels. Our algorithm is inspired by the recent progress of non‐local image processing techniques following the idea of ‘grouping and collaborative filtering’. In our framework, we first match and group similar patches in the input image, and then convert the problem of estimating missing values for the stack of matched patches to the problem of low‐rank matrix completion, and finally obtain the result by synthesizing all the restored patches. In our algorithm, how to accurately perform patch matching process and solve the low‐rank matrix completion problem are key points. For the first problem, we propose a robust patch matching approach, and for the second task, the alternating direction method of multipliers is employed. Experiments show that our algorithm has superior advantages over existing inpainting techniques. Besides, our algorithm can be easily extended to handle practical applications including rendering acceleration, photo restoration and object removal.     

基于加权核范数最小化的矩阵填充模型

协同过滤是目前推荐系统最常用的技术之一,相比于传统的推荐技术具有一定优势,但其缺点是受用户对商品评价的稀疏性制约,现阶段一般利用矩阵填充技术来解决这一问题。主要研究了基于低秩的矩阵填充模型,针对原有模型解对所有奇异值用同一值收缩的问题,提出了一种加权核范数最小化模型以提高核范数灵活度,给出了该模型用收缩算子可得到全局最优解的相关定理及证明,同时对模型的另一种形式在求解过程中的迭代收敛性进行了证明。用凸优化主流算法在两种真实数据集上进行的实验表明,改进后的模型一定程度上提高了计算速度与准确性。