拓扑梯度耦合FCMC的全自动图像修复优化算法

目的当前图像修复算法的损坏区域大都是依靠人工来确定,难以自动鉴定损坏区域,使其修复效率较低。此类算法通过利用像素缺失区域的间断边缘来完成填充,导致重构图像视觉间断,且都是依赖随机修复路径,增加了算法时耗。提出拓扑梯度最小重构路径耦合FCMC(Fuzzy C-mean Clustering)的全自动图像修复算法。方法基于图像损坏区域与完好区域之间的性质差异,引入模糊C均值(FCMC),通过损坏区域的聚类中心与各像素之间的距离来计算隶属度函数,设计基于FCMC的损坏区域自动鉴定算法,以自动识别待修复区域;再嵌入拓扑梯度,定义像素缺失区域的关键点选择规则,建立权重距离函数,得到像素缺失区域的连续轮廓,设计最低修复路径成本方案,完成图像重构;以PSNR(Peak Signal to Noise Ratio)为评估指标,构造图像修复反馈机制,优化修复图像。结果仿真结果显示:与当前图像修复算法相比,该算法可自动鉴定图像像素缺失区域,能够提取像素缺失区域的连续轮廓。同时,具有更好的修复视觉效果与更高的修复效率,重构图像不存在模糊与视觉不连通。结论提出的算法能够实现图像的全自动修复,可提高修复图像质量与效率。     

A modified POCS‐based reconstruction method for compressively sampled MR imaging

One of the challenging tasks in the application of compressed sensing to magnetic resonance imaging is the reconstruction algorithm that can faithfully recover the MR image from randomly undersampled k‐space data. The nonlinear recovery algorithms based on iterative shrinkage start with a single initial guess and use soft‐thresholding to recover the original MR image from the partial Fourier data. This article presents a novel method based on projection onto convex set (POCS) algorithm but it takes two images and then randomly combines them at each iteration to estimate the original MR image. The performance of the proposed method is validated using the original data taken from the MRI scanner at St. Mary's Hospital, London. The experimental results show that the proposed method can reconstruct the original MR image from variable density undersampling scheme in less number of iterations and exhibits better performance in terms of improved signal‐to‐noise ratio, artifact power, and correlation as compared to the reconstruction through low‐resolution and POCS algorithms. © 2014 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 24, 203–207, 2014     

Wavelet Based Restoration of Images with Missing or Damaged Pixels

This paper addresses the problem of how to restore degraded images where the pixels have been partly lost during transmission or damaged by impulsive noise. A wide range of image restoration tasks is covered in the mathematical model considered in this paper – e.g. image deblurring, image inpainting and super-resolution imaging. Based on the assumption that natural images are likely to have a sparse representation in a wavelet tight frame domain, we propose a regularization-based approach to recover degraded images, by enforcing the analysis-based sparsity prior of images in a tight frame domain. The resulting minimization problem can be solved efficiently by the split Bregman method. Numerical experiments on various image restoration tasks – simultaneously image deblurring and inpainting, super-resolution imaging and image deblurring under impulsive noise – demonstrated the effectiveness of our proposed algorithm. It proved robust to mis-detection errors of missing or damaged pixels, and compared favorably to existing algorithms.     

Image inpainting algorithm based on partial differential equation technique

Abstract

Image inpainting is the process of filling in missing parts of damaged images based on information gleaned from surrounding areas. In this paper, we present two variational models for image inpainting. Combining two models, we can simultaneously fill in missing, corrupted or undesirable information, while remove noise. We explain that diffusion performance of the proposed models is essentially superior to that of TV inpainting model by analysing the physical characteristics in local coordinates, and investigate the existence of minimising functionals in BV space. The experimental results show the effective performance of the proposed models in restoring scratched photos, text removal, and even removal of entire objects from images.     

基于引导因子与曲率惩罚模型的图像修复算法

目的 解决当前图像修复方法主要通过待修复像素点的法向量来确定修复过程,无法保证其修复顺序从破损区域的周边至中心进行,导致修复图像中存在块效应和不连续效应等问题。方法 将引导因子与曲率惩罚模型相结合,设计新的图像修复方法。利用破损区域的中心像素点与其他任意待修复像素点之间的距离来构造引导因子,并将其与置信度项以及数据项结合,形成优先权模型,用于选取优先修复块。利用待修复块的梯度特性对其平滑度进行判断,以明确该待修复块对应的最优匹配块的搜索范围,使其通过最小绝对差平方和（SSD）函数来搜索最优匹配块,从而将最优匹配块中像素点扩散填充至待修复块。最后,基于像素点间的等照度线曲率来建立曲率惩罚模型,以更新置信度项,从而实现图像的修复。结果 测试数据表明,与已有修复方案相比,所提算法可以更好地兼顾修复质量与效率。结论 所提方案具有较好的修复质量,可用于损坏面积较大图像的复原。     

The image inpainting algorithm based on pruning samples by referring to four-domains

In order to retrieve large scale damaged image with rich geometry structure and texture information, the novel image inpainting algorithm based on the neighbourhood reference priority can not only maintain image character but also improve image inpainting speed has been proposed in the paper. The problems of an image inpainting process can be translated into the best sample searching process. First, the extracting structure information of image and the dividing sample region into sub-regions server. Second, in order to adjust the neglect of structure of matching method named SSD, introducing structure symmetry matching constraint into matching method, it avoids matching mistakenly and improves sample matching precision and searching efficiency. Then, improving priority equations by bringing in structure weight and confidence, highlighted the effect of structure to inpainting sequence. Finally, computing priority of four-domains neighbour by computation overlapping information between object patch and sample patch, so that referring to secure information of four-domains neighbour, to prune sample dataset and search optimal sample. The experimental results have demonstrated that the proposed algorithm can overcome problems like texture blurring and structure dislocations and so on, the PSNR of the improved algorithm has been increased by 0.5–1?dB comparing other contrast methods while speeding up the image inpainting process, recovered image is much continuous for visuality. Meanwhile, it can recover efficiently common damage images and be more pervasive.     

基于Allen-Cahn方程图像修复的算子分裂方法（英）

本文提出了一种基于Allen-Cahn方程图像修复的算子分裂方法．其核心思想是利用算子分裂方法将原问题分解为一个线性方程和一个非线性方程,线性方程使用有限差分Crank-Nicolson格式进行离散,非线性方程利用解析方法进行求解,因此时间和空间都能达到二阶精度．由于该方法只作用于图像需要修复的区域,而其余区域的像素值与原始图像的保持一样,可以大大提高计算效率．合成图像和真实图像的数值实验验证了该算法的正确性和有效性．     

全变分自适应图像去噪模型

通过分析三种主要变分去噪模型(调和、全变分以及广义全变分模型)的优缺点,提出了一种基于全变分的自适应图像去噪模型。该模型根据噪声图像的信噪比,采用高斯滤波器对图像进行预处理,克服了全变分模型引入的阶梯效应;利用图像中每一像素点的梯度信息,自适应选取去噪模型中决定扩散强弱的参数p(x,y),使接近边缘处平滑较弱,远离边缘处平滑较强。数值实验表明,本方法在去除噪声的同时保留了图像的细节信息,取得了很好的降噪性能,其峰值信噪比(PSNR)在高噪声水平下,较其他变分方法至少提高1.0dB左右。     

基于调和修补模型的图像压缩方案

本文采用调和修补模型,给出了一种新的图像压缩方案。在图像编码端,采用边缘提取技术,仅对边缘扩展图像进行压缩编码,大大减少了需要编码的信息量。在图像解码端,采用调和修补模型来重建原始图像。通过对大量灰度和彩色图像进行实验,结果表明,该压缩方案能够在较少信息量的情况下得到很好的重建图像。     

A novel image inpainting technique based on median diffusion

Image inpainting is the technique of filling-in the missing regions and removing unwanted objects from an image by diffusing the pixel information from the neighbourhood pixels. Image inpainting techniques are in use over a long time for various applications like removal of scratches, restoring damaged/missing portions or removal of objects from the images, etc. In this study, we present a simple, yet unexplored (digital) image inpainting technique using median filter, one of the most popular nonlinear (order statistics) filters. The median is maximum likelihood estimate of location for the Laplacian distribution. Hence, the proposed algorithm diffuses median value of pixels from the exterior area into the inner area to be inpainted. The median filter preserves the edge which is an important property needed to inpaint edges. This technique is stable. Experimental results show remarkable improvements and works for homogeneous as well as heterogeneous background. PSNR (quantitative assessment) is used to compare inpainting results.     

基于相似稀疏度与局部差异特征的图像修复算法

目的为了解决当前图像修复算法在破损面积较大时,其复原图像易丢失局部细节信息而导致修复图像存在振铃效应以及不连贯效应的不足,提出一种基于相似稀疏度耦合局部差异特征的图像修复算法。方法首先,利用待修复块及其相邻块内像素的均方距离来构造相似稀疏度模型,以形成优先权度量函数,根据其计算的优先级来确定优先修复块。然后,通过样本块对应的梯度向量模值来构造局部差异因子,计算样本块的局部差异,并以计算结果为依据对样本块的尺寸进行调整。最后,以像素点的颜色差值信息为依据,构造近似函数,选取最优匹配块对待修复块进行复原。结果仿真实验结果显示,与当前图像修复算法相比,该算法具有更高的修复质量和效率,其复原图像不存在振铃效应和不连续效应等。结论所提算法具有较高的修复视觉质量,能用于大面积损坏图像的复原。     

一种改进的视频序列超分辨率重建算法及应用

在凸集优化基础上,充分利用最大后验概率和凸集投影技术,提出了一种高效强鲁棒性视频序列分辨率提升算法。首先,在空域设计一个简单的预处理共轭梯度估计器,预测原始高分辨率图像;然后,在小波域分别创建帧间和帧内两个不同的凸集,并实施不同的投影运算,提取出隐含在相邻低分辨率图像中的细节信息;最后,利用空域估计器中相邻因子间的关系约束凸集投影解的可行域,保证快速获得图像重建的唯一最优解。仿真实验和实际交通监测系统应用结果均表明,该方法较其他方法不仅可获得更高的峰值信噪比和更好的可视化效果,而且收敛更快,鲁棒性更强。     

基于局部坐标二次微分的自适应全变分去噪复原

针对传统TV去噪复原算法以梯度模值作为图像的边缘检测算子,无法清晰地识别边缘和灰度渐变区及去除平坦区内的孤立噪声的问题,提出了一种基于局部坐标二次微分的边缘检测算子对传统模型进行改进。改进后的模型能根据各像素点的新检测算子信息,自适应选取复原模型中决定扩散强弱的参数,并且利用图像局部信息对正则化项和保真项进行加权。同时在数值实现上,采用一种基于梯度矢量的方向变化的方法来实现散度离散化,以更加有效地保留图像的局部细节信息。数值试验表明,该算法在克服灰度渐变区内的阶梯效应和保留图像的细节边缘方面明显优于传统算法。     

基于平滑度测量与互相关制约的图像修复算法

目的为了解决当前较多图像修复算法在对损坏区域实施填充修复时没有考虑图像块之间的相关性,导致修复图像中存在振铃以及连接间断等不足,拟设计平滑度测量因子耦合互相关制约的图像修复算法。方法将图像的Laplace算子引入到优先权的计算过程中,以增加图像的边缘信息,优化数据项的成分,并利用置信度、图像的Laplace算子和数据项构造优先权计算模型,以获取优先修复块。利用等照度线的二阶导数,构造平滑度测量因子获取图像的平滑度,并以图像的平滑度为导向,对最优匹配块的搜索区域进行定位。通过误差平方和函数(SumofSquaredDifferences,SSD)在定位的搜索区域中搜寻最优匹配块,并采用互相关系数函数对最优匹配块的唯一性进行制约,以提高所获最优匹配块的准确性。结果实验结果显示,与当前修复技术相比,所提方法具有更高的连接完整性和清晰度。结论所提方案可对损坏图像完成较好的视觉复原,在图像信息处理领域具有一定的参考价值。     

Video object inpainting: a scale-robust method

Abstract

Video inpainting is the process of reconstructing damaged regions of corrupted frames. In this research, we raise a few issues in existing video inpainting systems. They are usually not robust to the change in the object scale and cannot handle large missing regions behind the moving object. In this attempt, we will address the above issues as following: first, we extract moving objects from the background and construct two mosaic images for each object, a small mosaic and a large mosaic image. The small mosaic is used to detect the amount of scale changes in the moving objects and the large one is used to inpaint partially or completely corrupted objects. We next place the inpainted moving foreground in its location and rescale the objects to their original scale. Finally, we combine the inpainted moving foreground and the background to obtain the corrected video. To speed up the process, we have utilised a multi-resolution approach so that the patch are initially matched in a coarse resolution and later are refined in a fine resolution. The results confirm the robustness of our method in handling the scale change of moving objects and large missing regions.     

多级FFD配准视频人脸超分辨率重建

超分辨率重建是解决视频人脸识别中人脸分辨率低的有效方法,但由于人脸畸变、表情变化等非刚性变化导致无法精确配准和重建.针对此问题,提出基于B样条的多级模型自由形式形变(FFD)弹性配准算法.先用低分辨率FFD网格全局配准,再对全局配准后的图像分块并计算对应子图块的相关性系数,对相关性系数小的子图块用高分辨率FFD网格局部细配准.在配准的寻优过程中采用基于混沌因子的自适应步长最速下降法提高寻优效率.配准后,采用POCS算法对多帧图像重建高分辨率图像来识别.在标准视频库和自建视频库上实验仿真,结果表明在人脸畸变和表情变化很大的情况下,也能够精确的配准和很好的重建,得到较高识别率.     

基于匹配调节法则和梯度约束模型的图像修复算法

目的针对当前较多图像修复算法难以根据不同纹理结构来自适应调整修复块的尺寸,导致修复结果中存在不连续效应和模糊效应等不足,提出一种结合匹配调节法则和梯度约束模型的图像修复算法。方法首先,利用平滑因子对置信度项进行约束,构造优先级判定模型,对待修复块的优先级进行度量,确定优先修复块。随后,通过SSD模型度量样本块之间的匹配结果,并根据匹配结果制定匹配调节法则,使得样本块能根据匹配度自适应调节其大小,以提高修复质量。最后,将梯度模值块中像素点的均方距离度量结果与样本块中像素点相结合,构造梯度约束模型,用以获取最佳匹配块对待修复块进行填充修复。结果实验结果表明,与当前图像修复算法相比,该算法修复的图像具有更好的修复质量,在像素丢失率较高的情况下,仍然具有较高的相似度值。结论所提算法具备较好的修复视觉质量,可用于被大面积损坏图像的修复。     

A comparison of Lanczos and optimization methods in the partial solution of sparse symmetric eigenproblems

In the present paper, we analyse the computational performance of the Lanczos method and a recent optimization technique for the calculation of the p (p ≤ 40) leftmost eigenpairs of generalized symmetric eigenproblems arising from the finite element integration of elliptic PDEs. The accelerated conjugate gradient method is used to minimize successive Rayleigh quotients defined in deflated subspaces of decreasing size. The pointwise Lanczos scheme is employed in combination with both the Cholesky factorization of the stiffness matrix and the preconditioned conjugate gradient method for evaluating the recursive Lanczos vectors. The three algorithms are applied to five sample problems of varying size up to almost 5000. The numerical results show that the Lanczos approach with Cholesky triangularization is generally faster (up to a factor of 5) for small to moderately large matrices, while the optimization method is superior for large problems in terms of both storage requirement and CPU time. In the large case, the Lanczos–Cholesky scheme may be very expensive to run even on modern quite powerful computers.     

UFC-Net with Fully-Connected Layers and Hadamard Identity Skip Connection for Image Inpainting

Image inpainting is an interesting technique in computer vision and artificial intelligence for plausibly filling in blank areas of an image by referring to their surrounding areas. Although its performance has been improved significantly using diverse convolutional neural network (CNN)-based models, these models have difficulty filling in some erased areas due to the kernel size of the CNN. If the kernel size is too narrow for the blank area, the models cannot consider the entire surrounding area, only partial areas or none at all. This issue leads to typical problems of inpainting, such as pixel reconstruction failure and unintended filling. To alleviate this, in this paper, we propose a novel inpainting model called UFC-net that reinforces two components in U-net. The first component is the latent networks in the middle of U-net to consider the entire surrounding area. The second component is the Hadamard identity skip connection to improve the attention of the inpainting model on the blank areas and reduce computational cost. We performed extensive comparisons with other inpainting models using the Places2 dataset to evaluate the effectiveness of the proposed scheme. We report some of the results.     

Efficient edge-directed deblocking algorithm for KOMPSAT-2 satellite images

In this paper, we present a deblocking algorithm that removes grid and staircase artefacts, which result from the compression of images. The KORea Multi-Purpose SATellite-2 images are efficiently compressed using an algorithm that assigns quantisation coefficients, by rows, according to local region complexity, in such a way that more complicated regions are more compressed. However, this approach is problematic in that it produces blocking artefacts in those relatively less complicated areas within the same row as the complicated regions. Removing these artefacts with a general deblocking algorithm can, undesirably, blur the complicated regions as well as the artefacts; additionally, the general filter is poor at preserving curved edges. Therefore, the algorithm proposed herein presents an adaptive filtering scheme for removing blocking artefacts while preserving image details, including curved edges, using a content analysis. Particularly, edge adaptive filtering has been introduced to remove staircase artefacts around the big image edges. Experimental analyses revealed that the proposed method outperforms several state-of-the-art approaches in terms of subjective or objective image quality, and computational speed.