结合边缘信息和门卷积的人脸修复算法

针对任意形状遮挡下人脸修复,现有方法容易产生边缘模糊和恢复结果失真等问题。提出了一种结合边缘信息和门卷积的人脸修复算法。首先,通过先验人脸知识产生遮挡区域的边缘图,以约束人脸修复过程。其次,利用门卷积在部分像素缺失下的精确局部特征描述能力,设计面向图像修复的门卷积深度生成对抗网络(GAN)。该模型由边缘连接生成对抗网络和图像修复生成对抗网络两部分组成。边缘连接网络利用二值遮挡图和待修复图像及其边缘图的多源信息进行训练,实现对缺失边缘图像的自动补全和连接。图像修复网络以补全的边缘图为引导信息,联合遮挡图像进行缺失区域修复。实验结果表明:相比其他算法,该算法修复效果更好,其评价指标比当前基于深度学习的图像修复算法更优。     

基于各向异性插值模型的快速图像修复方法*

基于偏微分方程的图像修复算法运行缓慢,且无法恢复纹理细节,实用性较差。基于地统计学思想,提出一种简单有效的基于各向异性插值模型的图像修复方法。实验表明该方法具有计算复杂度低和能够恢复图像纹理细节的优点,对于图像小区域划痕具有很好的修复效果,具有较高的实用价值。     

Example-based 3D inpainting of point clouds using metric tensor and Christoffel symbols

In this paper, we address the problem of 3D inpainting using example-based methods for point cloud data. 3D inpainting is a process of filling holes or missing regions in the reconstructed 3D models. Typically inpainting methods addressed in the literature fill missing regions due to occlusions or inaccurate scanning of 3D models. However, we focus on scenarios involving naturally existing damaged models which are partly broken or incomplete in artifacts at cultural heritage sites. We propose two example-based inpainting techniques, namely region of interest (ROI)-based and patch-based methods, to inpaint the missing regions of the damaged model. For both the methods, we represent the 3D model as a set of Riemannian manifolds in Euclidean space, to capture the inherent geometry using metric tensor and Christoffel symbols as geometric features and decompose into basic shape (such as spherical, conical and cylindrical) regions using decomposition algorithm derived from supervised learning. In ROI-based method, instead of using single similar example for inpainting, we select the most relevant regions that best-fit the missing region from the set of basic shape regions derived from n similar examples. And in patch-based method, we not only select the most relevant regions but cluster the regions into a set of patches. The best corresponding patches that match the missing region to be inpainted are considered to be the most relevant best-fit patches that cover the complete missing region. We demonstrate the performance of proposed inpainting methods on cultural heritage artifacts with varying complexities and sizes for both synthetically generated holes and real missing regions.     

划分特征子区域的图像修复算法

为了解决含有丰富纹理信息和复杂结构信息的大破损区域中的缺失信息修复的问题,提出了一种划分特征子区域的图像修复算法。首先,根据图像中包含的不同特征,运用特征公式进行特征提取,再通过统计特征值划分特征子区域,提高了图像修复的速度;其次,在原Criminisi算法的基础上改进了优先级的计算,通过增大结构项的影响,避免结构断裂的产生;然后,通过目标块和其最佳邻域相似块共同约束样本块的选取,确定最佳样本块集;最后,利用权值分配法合成最佳样本块。实验结果表明,所提算法相比原Criminisi算法,其峰值信噪比（PSNR）提升了2~3 dB,相比基于稀疏表示的块优先权值计算的算法,其修复效率有明显的提高。所提算法不但适用于一般小尺度的破损图像的修复,而且对于含有丰富纹理信息和复杂结构信息的大破损图像的修复效果也更佳,并且修复后的图像更加符合人们视觉上的连通性。     

Automatic inpainting by removing fence-like structures in RGBD images

Recent inpainting techniques usually require human interactions which are labor intensive and dependent on the user experiences. In this paper, we introduce an automatic inpainting technique to remove undesired fence-like structures from images. Specifically, the proposed technique works on the RGBD images which have recently become cheaper and easier to obtain using the Microsoft Kinect. The basic idea is to segment and remove the undesired fence-like structures by using both depth and color information, and then adapt an existing inpainting algorithm to fill the holes resulting from the structure removal. We found that it is difficult to achieve a satisfactory segmentation of such structures by only using the depth channel. In this paper, we use the depth information to help identify a set of foreground and background strokes, with which we apply a graph-cut algorithm on the color channels to obtain a more accurate segmentation for inpainting. We demonstrate the effectiveness of the proposed technique by experiments on a set of Kinect images.     

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...     

一范数在图像修复中的应用

考虑到图像相邻像素具有相关性,修复后的图像能量应为最低状态。定义了复合向量的范数。利用改进的热传导模型进行图像修复。提出的算法既可用来有效地修复划痕,也可用于去除文字,还可以填充较大破损区域。     

An improved image quality algorithm for exemplar-based image inpainting

Image inpainting is a common technique for repairing image regions that are scratched or damaged. This process involves reconstructing damaged parts and filling-in regions in which data/colour information is missing. There are many potential applications for image inpainting, such as repairing old images, repairing scratched images, removing unwanted objects, and filling-in missing areas. This paper develops an exemplar-based algorithm, one of the most important and popular image inpainting techniques, to fill-in missing regions caused by removing unwanted objects, image compression, scratches, or image transformation via the Internet. The proposed algorithm includes two phases of searching to select the best-matching information. In the first phase, the searching mechanism uses the entire image to find and select the most similar patches using the Euclidean distance. The second phase measures the distance between the location of the selected patches and the location of the patch to be filled. The performance of the proposed approach is evaluated through comprehensive experiments on several well-known images used in this area of research. The experimental results demonstrate the superior performance of the proposed approach over some state-of-the-art approaches in terms of quality in terms of both objective (using the peak signal-to-noise ratio (PSNR) as well as the structural similarity index method (SSIM)) and subjective (i.e., visual) measures.

     

Image inpainting with salient structure completion and texture propagation

Image inpainting technique uses structural and textural information to repair or fill missing regions of a picture. Inspired by human visual characteristics, we introduce a new image inpainting approach which includes salient structure completion and texture propagation. In the salient structure completion step, incomplete salient structures are detected using wavelet transform, and completion order is determined through color texture and curvature features around the incomplete salient structures. Afterwards, curve fitting and extension are used to complete the incomplete salient structures. In the texture propagation step, the proposed approach first synthesizes texture information of completed salient structures. Then, the texture information is propagated into the remaining missing regions. A number of examples on real and synthetic images demonstrate the effectiveness of our algorithm in removing occluding objects. Our results compare favorably to those obtained by existing greedy inpainting techniques.     

一种双判别器GAN的古彝文字符修复方法EI北大核心CSCD

在中国,彝文古籍文献日益流失而且损毁严重,由于通晓古彝文的研究人员缺乏,使得古籍恢复工作进展十分缓慢.人工智能在图像文本领域的应用,为古籍文献的自动修复提供可能.本文设计了一种双判别器生成对抗网络(Generative adversarial networks with dual discriminator,D2GAN),以还原古代彝族字符中的缺失部分.D2GAN是在深度卷积生成对抗网络的基础上,增加一个古彝文筛选判别器.通过三个阶段的训练来迭代地优化古彝文字符生成网络,以获得古彝文字符的文字生成器.根据筛选判别器的损失结果优化D2GAN模型,并使用生成的字符恢复古彝文中丢失的笔画.实验结果表明,在字符残缺低于1/3的情况下,本文提出的方法可使文字笔画的修复率达到77.3%,有效地加快了古彝文字符修复工作的进程.     

一种改进的基于样图的图像修复算法

图像修复的目的是填补有信息缺损的图像,并使观察者无法察觉出图像的填补痕 迹。分析了图像修复技术中的Criminisi算法,针对它的不足,提出一种新的改进算法。新算法从4个方面加以改进：使用了新的优先权计算函数,优化了优先权大小选择的计算,避免了因模板数据项迅速衰减带来的错误填充次序;利用Sobel算子改进等照度线计算,使等照度线上的点优先被修复;采用新的匹配方法,将查找匹配的范围锁定在破损区域的边缘。最后为了平滑置信值更新导致的误差传播,定义了新的置信值更新方程。实验结果证明,本文图像修复算法不但可以改善图像修复质量,还可以提高图像修复效率。     

基于生成对抗网络的图像修复算法

为解决当前基于生成对抗网络的深度学习网络模型在面对较复杂的特征时存在伪影、纹理细节退化等现象, 造成视觉上的欠缺问题, 提出了连贯语义注意力机制与生成对抗网络相结合的图像修复改进算法. 首先, 生成器使用两阶段修复方法, 用门控卷积替代生成对抗网络的普通卷积, 引入残差块解决梯度消失问题, 同时引入连贯语义注意力机制提升生成器对图像中重要信息和结构的关注度; 其次, 判别器使用马尔可夫判别器, 强化网络的判别效果, 将生成器输出结果进行反卷积操作得到最终修复后的图片. 通过修复结果以及图像质量评价指标与基线算法进行对比, 实验结果表明, 该算法对缺失部分进行了更好地预测, 修复效果有了更好的提升.     

基于分组行进算法的图像修补方法

图像修补是图像恢复研究中的一个重要内容,它的目的是根据图像的现有信息来自动恢复丢失的信息。虽然图像修补的基本思想十分简单,但是许多的图像修补算法都十分复杂,而且难于实现。快速行进算法(FMM)与水平集法(Level Set)相结合进行曲线进化是一种高效的曲线进化算法,该算法的时间复杂度是O(NlbN)。Kim提出了另一种水平集的曲线进化算法——分组行进算法(GMM),该算法的时间复杂度是O(N)。受其启发,为了更快地进行图像修补,提出了一种基于GMM算法的图像修补的新算法,并研究了对GMM算法的细节改进。为了验证算法的快速性,还给出了使用Bertalmio提出的算法、Telea提出的算法以及新算法对同一幅图片进行修补的实验结果。通过比较发现,该新算法在大幅度提高修补速度的同时,仍能保持较好的修补效果。     

基于变量分离和交替最小化的图像修复

提出变量分离和交替最小化相结合解决l1正则优化问题,并用于非纹理图像的修复。基于变量分离技术,该算法将目标函数的l1成分和l2成分解耦,l1正则优化问题简化为一系列非约束优化问题。除了交替最小化迭代地求解这些非约束优化问题,还引入投影法加快和简化求解过程。实验在有噪声和无噪声的情况下,用提出的算法对信息丢失30%的图像进行修复。实验结果表明：该算法可有效解决包括图像修复在内的一系列图像复原问题;与某些同类算法相比,在修复速度和修复效果方面均具有优势。     

Optimum inpainting for depth map based on L 0 total variation

Depth map generated by the Kinect may have some pixels lost due to echo attenuation of infra- red light and mutual interference between neighboring pixels, which can cause pervasive problems when utilizing Kinect cameras as depth sensors. In this work, we propose a 2-step inpainting algorithm to infill the holes. First, a naive Bayesian estimation is conducted as preliminary inpainting scheme, utilizing neighboring pixels of the missing ones, and corresponding pixels in the color image as prior knowledge. After that, an optimization is implemented to improve the depth map, where the false edges in mistakenly inpainted regions are detected, then iteratively propelled to their true positions under total variation framework. Experimental results are included to show effectiveness of the proposed algorithm.     

采用加权优化的图像修复

针对目前贪婪修复算法可能存在修复效果视觉不一致以及优化修复算法中存在的算法复杂度较高或者未考虑结构信息的情况,提出一种基于加权优化的图像修复算法,通过定义出新的能量函数,把图像破损修复问题转化为加权的离散优化问题,在保证结构信息强、信任度高的区域被优先修复的前提下,利用贪婪修复思想获取初值并计算权值,然后通过类EM算法迭代求解出破损区域中每一个像素的最佳值。与其他贪婪合成和最优化方法相比,优先考虑结构信息对修复效果的影响,更好地保持了纹理和结构的整体一致性。     

基于语义先验和双通道特征提取的图像修复

针对现有的图像修复算法重建结果存在的局部结构不连通、细节还原不准确等问题,提出了一种基于语义先验和双通道特征提取的图像修复算法（semantic prior and dual channel extraction,SPDCE）。该算法利用语义先验网络学习缺失区域的语义信息和上下文知识,对缺失区域进行预测,增强了生成图像的局部一致性;然后通过双通道特征提取网络充分挖掘图像信息,提升了对纹理细节的感知和利用能力;再使用上下文特征调整模块在多个尺度上捕获并编码丰富的语义特征,从而生成更真实的图像视图和更精细的纹理细节。在CelebA-HQ和Places2数据集上进行实验验证,结果表明,SPDCE算法与常用算法相比,峰值信号比（peak signal-to-noise ratio,PSNR）和结构相似性（structural similarity,SSIM）分别提升1.6~1.73 dB和3.1%~9.9%,L1 loss下降15.2%~27.8%。实验证明所提算法修复后的图像具有更合理的结构和更丰富的细节,图像修复效果更优。     

基于冗余字典学习的图像修补算法

在稀疏表示理论研究的基础上,提出了基于不同冗余字典的图像修补算法。首先设计采用离散余弦变换或K-SVD算法获得冗余DCT字典、KSVDG全局字典及KSVDA自适应字典等三种不同的字典;然后分别基于上述三种不同的冗余字典,稀疏表示待处理图像;最终图像中缺损的部分将通过冗余字典和稀疏系数有效地表示出来。实验结果表明,提出的算法修补后的图像视觉效果好,并在峰值信噪比、特征相似度等主要图像质量评价指标上优于现有几种经典的图像修补方法。     

基于全变分小波域缺损修复算法的一种数值实现

针对基于全变分(TV)小波域图像修复算法,提出了一种新的数值计算方法。算法充分利用了图像待修复像素点周围八邻域的信息计算曲率在像素域的近似值。与传统近似求解方法不同,该数值近似计算方法不仅保证了近似结果具有更高的精确性,而且对噪声的鲁棒性好。对不同丢失率的图像用所提出的数值计算方法取得了较好的修复效果,尤其当小波系数丢失率较高时效果更为明显。所提方法可以为图像压缩导致的系数缺损修复提供解决思路。