Feature matching based on unsupervised manifold alignment

Feature-based methods for image registration frequently encounter the correspondence problem. In this paper, we formulate feature-based image registration as a manifold alignment problem, and present a novel matching method for finding the correspondences among different images containing the same object. Different from the semi-supervised manifold alignment, our methods map the data sets to the underlying common manifold without using correspondence information. An iterative multiplicative updating algorithm is proposed to optimize the objective, and its convergence is guaranteed theoretically. The proposed approach has been tested for matching accuracy, and robustness to outliers. Its performance on synthetic and real images is compared with the state-of-the-art reference algorithms.     

Unsupervised manifold alignment using soft-assign technique

In this paper, we propose a robust unsupervised algorithm for automatic alignment of two manifolds in different datasets with possibly different dimensionalities. The significant contribution is that the proposed alignment algorithm is performed automatically without any assumptions on the correspondences between the two manifolds. For such purpose, we first automatically extract local feature histograms at each point of the manifolds and establish an initial similarity between the two datasets by matching their histogram-based features. Based on such similarity, an embedding space is estimated where the distance between the two manifolds is minimized while maximally retaining the original structure of the manifolds. The elegance of this idea is that such complicated problem is formulated as a generalized eigenvalue problem, which can be easily solved. The alignment process is achieved by iteratively increasing the sparsity of correspondence matrix until the two manifolds are correctly aligned and consequently one can reveal their joint structure. We demonstrate the effectiveness of our algorithm on different datasets by aligning protein structures, 3D face models and facial images of different subjects under pose and lighting variations. Finally, we also compare with a state-of-the-art algorithm and the results show the superiority of the proposed manifold alignment in terms of vision effect and numerical accuracy.     

一种基于全局和局部特征匹配的流形对齐算法

不同流形样本点之间的关联性挖掘是决定流形对齐算法效率的关键问题。提出了一种新的思路,利用测地距离初步构造不同流形样本点之间的关联性,再利用样本点之间局部几何结构的相似性进行修正,以更为准确地挖掘不同流形样本点之间的关联性。进一步提出一种新的半监督流形对齐算法,利用已知对应点信息和所挖掘样本点之间的关联性,将多个流形数据投影到共同的低维空间。与传统的半监督流形对齐算法相比,本算法在先验信息不充分的情况下,能更准确地联结不同流形数据集。最后通过在实际数据集上的实验验证了算法的有效性。     

基于空间约束的正则化流形学习影像匹配方法

针对大数据量的影像匹配问题,提出了一种基于正则化流形学习的影像匹配方法。该方法利用曼哈顿距离方法计算特征点SIFT描述符的相似性,并在此基础上增加同一幅影像中特征点之间的空间结构关系作为正则化项,采用流形学习方法将两幅影像中的特征点共同映射到同一流形空间;根据最小距离方法进行特征点的匹配。通过选取四种不同来源的影像对进行实验,与SVD-SIFT方法、LE-SIFT方法进行综合分析,结果表明该方法在匹配性能上优于现有方法,而且该方法具有线性复杂度,适用于处理特征点数量较大的影像匹配。     

基于双序列比对算法的立体图像匹配方法*

在分析现有立体匹配方法的基础上,提出一种基于双序列比对算法的立体图像匹配方法。将立体图像对中同名极线上的像素灰度值看做是一对字符序列,使用基于动态规划思想的双序列比对算法对这些对字符序列进行匹配,以获取立体图像视差。为验证该方法的可行性和适用性,采用人脸立体图像对进行实验。实验结果表明,使用该方法进行立体图像匹配能获得光滑的、稠密的视差图。基于动态规划思想的双序列比对算法能够有效地解决立体图像匹配问题,从而为图像的立体匹配提供了一个实用有效的方法。     

Unsupervised group matching with application to cross-lingual topic matching without alignment information

We propose a method for unsupervised group matching, which is the task of finding correspondence between groups across different domains without cross-domain similarity measurements or paired data. For example, the proposed method can find matching of topic categories in different languages without alignment information. The proposed method interprets a group as a probability distribution, which enables us to handle uncertainty in a limited amount of data, and to incorporate the high order information on groups. Groups are matched by maximizing the dependence between distributions, in which we use the Hilbert Schmidt independence criterion for measuring the dependence. By using kernel embedding which maps distributions into a reproducing kernel Hilbert space, we can calculate the dependence between distributions without density estimation. In the experiments, we demonstrate the effectiveness of the proposed method using synthetic and real data sets including an application to cross-lingual topic matching.     

Unsupervised manifold learning based on multiple feature spaces

Manifold learning is a well-known dimensionality reduction scheme which can detect intrinsic low-dimensional structures in non-linear high-dimensional data. It has been recently widely employed in data analysis, pattern recognition, and machine learning applications. Isomap is one of the most promising manifold learning algorithms, which extends metric multi-dimensional scaling by using approximate geodesic distance. However, when Isomap is conducted on real-world applications, it may have some difficulties in dealing with noisy data. Although many applications represent a special sample by multiple feature vectors in different spaces, Isomap employs samples in unique observation space. In this paper, two extended versions of Isomap to multiple feature spaces problem, namely fusion of dissimilarities and fusion of geodesic distances, are presented. We have employed the advantages of several spaces and depicted the Euclidean distance on learned manifold that is more compatible to the semantic distance. To show the effectiveness and validity of the proposed method, some experiments have been carried out on the application of shape analysis on MPEG7 CE Part B and Fish data sets.     

基于图像蒙板的无监督图像风格迁移

目前大多数的图像风格迁移方法属于有监督学习,训练数据需要成对出现,并且在处理图像背景时,现有的方法过于繁琐。针对这些问题,提出了一种基于图像蒙板的无监督图像风格迁移方法。在实验中,采用了基于循环一致性的CycleGAN架构,并使用Inception-ResNet结构设计了一个全新的具有内置图像蒙板的生成式模型,最后通过无监督学习将图像的背景与学习到的抽象特征进行自动重组。实验表明,新方法有效地对图像背景和抽象特征进行自动分离与重组,同时解决了特征学习过程中的区域干扰问题,获得了可观的视觉效果。     

黎曼流形上的保局投影在图像集匹配中的应用

目的提出了黎曼流形上局部结构特征保持的图像集匹配方法。方法该方法使用协方差矩阵建模图像集合,利用对称正定的非奇异协方差矩阵构成黎曼流形上的子空间,将图像集的匹配转化为流形上的点的匹配问题。通过基于协方差矩阵度量学习的核函数将黎曼流形上的协方差矩阵映射到欧几里德空间。不同于其他方法黎曼流形上的鉴别分析方法,考虑到样本分布的局部几何结构,引入了黎曼流形上局部保持的图像集鉴别分析方法,保持样本分布的局部邻域结构的同时提升样本的可分性。结果在基于图像集合的对象识别任务上测试了本文算法,在ETH80和YouTube Celebrities数据库分别进行了对象识别和人脸识别实验,分别达到91.5%和65.31%的识别率。结论实验结果表明,该方法取得了优于其他图像集匹配算法的效果。     

基于属性论的无监督图像聚类

提出了应用属性论来实现无监督图像聚类的方法。首先提取图像的底层特征，再利用属性论方法中定性映射、转换程度函数理论对图像进行分类。由于一次聚类并未达到很好的效果，因此采取了多次聚类的策略。实验结果表明，本方法简单明了并且有效。     

稀疏回归和流形学习的无监督特征选择算法

针对无标签高维数据的大量出现,对机器学习中无监督特征选择进行了研究。提出了一种结合自表示相似矩阵和流形学习的无监督特征选择算法。首先,通过数据的自表示性质,构建相似矩阵,结合低维流形能够表示高维数据结构这一流形学习思想,建立一种考虑流形学习的无监督特征选择优化模型。其次,为了保证选择更有用及更稀疏的特征,采用◢l◣▼2,1▽范数对优化模型进行约束,使特征之间相互竞争,消除冗余。进而,通过变量交替迭代对优化模型进行求解,并证明了算法的收敛性。最后,通过与其他几个无监督特征算法在四个数据集上的对比实验,表明所给算法的有效性。     

基于目标图像先验信息的无监督多聚焦图像融合

多聚焦图像融合（MFIF）是从不同源图像中获取聚焦区域,以形成全清晰图像的一种图像增强方法。针对目前MFIF方法主要存在的两个方面问题,即传统的空间域方法在其融合边界存在较强的散焦扩散效应（DSE）以及伪影等问题;深度学习方法缺乏还原光场相机生成的数据集,并且因需要大量手动调参而存在训练过程耗时过多等问题,提出了一种基于目标图像先验信息的无监督多聚焦图像融合方法。首先,将源图像本身的内部先验信息和由空间域方法生成的初始融合图像所具有的外部先验信息分别用于G-Net和F-Net网络输入,其中,G-Net和F-Net网络都是由UNet组成的深度图像先验（DIP）网络;然后,引入一种由空间域方法生成的参考掩膜辅助G-Net网络生成引导决策图;最后,该决策图联合初始融合图像对F-Net网络进行优化,并生成最终的融合图像。验证实验基于具有真实参考图像的Lytro数据集和融合边界具有强DSE的MFFW数据集,并选用了5个广泛应用的客观指标进行性能评价。实验结果表明,该方法有效地缩短了优化迭代次数,在主观和客观性能评价上优于8种目前最先进的MFIF方法,尤其在融合边界具有强DSE的数据集上表现得更有优势。     

流形学习与基于线性耦合映射的流形对齐

从一些具有代表性的经典流形学习方法的回顾来看,传统的流形学习主要处理来自单一流形的数据的降维问题.随着流形学习研究的不断深入,以多流形作为研究对象的流形学习问题逐步引起了研究者的注意.提出了一种基于线性耦合映射的流形对齐算法.算法克服非线性流形对齐算法不能够直接处理Out-of-sample数据的问题.同时,与已有的线性流形对齐算法相比,该算法不需要假设流形间满足仿射变换关系,因而能够更加灵活地处理一些比较实际的流形对齐问题.     

基于相关度距离的无监督并行哈希图像检索

针对传统无监督哈希图像检索模型中存在图像数据之间的语义信息学习不足,以及哈希编码长度每换一次模型就需重新训练的问题,提出一种用于大规模图像数据集检索的无监督搜索框架——基于相关度距离的无监督并行哈希图像检索模型.首先,使用卷积神经网络(CNN)学习图像的高维特征连续变量;然后,使用相关度距离衡量特征变量构建伪标签矩阵,...     

Unsupervised image retrieval framework based on rule base system

This paper introduces unsupervised image retrieval framework based on a rule base system. The proposed framework makes use of geometric moments (GMs) for features extraction. The main advantage with the GMs is that image coordinate transformations can be easily expressed and analyzed in terms of the corresponding transformations in the moment space. These features are used to perform the image mining for acquiring clustering knowledge from a large empirical images database. Irrelevance between images of the same cluster is precisely considered in the proposed framework through a relevant feedback phase followed by a novel clustering refinement model. The images and their corresponding classes pass to a rule base algorithm for extracting a set of accurate rules. These rules are pruning and may reduce the dimensionality of the extracted features. The advantage of the proposed framework is reflected in the retrieval process, which is limited to the images in the class of rule matched with the query image features. Experiments show that the proposed model achieves a very good performance in terms of the average precision, recall and retrieval time compared with other models.     

Unsupervised image segmentation based on MRFs and graph cuts

Markov random fields （MRFs） can be used for a wide variety of vision problems. In this paper we will propose a Markov random field （MRF） image segmentation model. The theoretical framework is based on Bayesian estimation via the energy optimization. Graph cuts have emerged as a powerful optimization technique for minimizing energy functions that arise in low-level vision problem. The theorem of Ford and Fulkerson states that min-cut and max-flow problems are equivalent. So, the minimum s/t cut problem can be solved by finding a maximum flow from the source s to the sink t. we adopt a new min-cut/max-flow algorithm which belongs to the group of algorithms based on augmenting paths. We propose a parameter estimation method using expectation maximization （EM） algorithm. We also choose Gaussian mixture model as our image model and model the density associated with one of image segments （or classes） as a multivariate Gaussian distribution. Characteristic features related to the information in color, texture and position are extracted for each pixel. Experimental results will be provided to illustrate the performance of our method.     

基于生成对抗网络的无监督图像风格迁移

无监督的图像风格迁移是计算机视觉领域中一个非常重要且具有挑战性的问题.无监督的图像风格迁移旨在通过给定类的图像映射到其他类的类似图像.一般情况下成对匹配的数据集很难获得,这极大限制了图像风格迁移的转换模型.因此,为了避免这种限制,对现有的无监督的图像风格迁移的方法进行改进,采用改进的循环一致性对抗网络进行无监督图像风格迁移.首先为了提升网络的训练速度,避免梯度消失的现象出现,在传统的循环一致性网络生成器部分引入DenseNet网络;在提高生成器的性能方面,生成器网络部分引入attention机制来输出效果更好的图像;为了减少网络的结构风险,在网络的每一个卷积层都使用谱归一化.为了验证本文方法的有效性,在monet2photo、vangogh2photo和facades数据集上进行了实验,实验结果表明,该方法在Inception score平均分数和FID距离评价指标上均有所提高.     

Method for fingerprint minutiae matching based on their alignment

In the paper a method for fingerprint matching is presented, which is based on the use of additional information about the representation of fingerprint minutiae. The method proposed is a modification of the method by M. Tico et al [1]. However, in contrast with the prototype, this method uses both the characteristic of local regions of the directional field and the local configurations of minutiae present in the fingerprint. Vladimir Nikolaevich Perminov. Graduated from the Moscow State Institute of Electronic Technology. Candidate of Engineering. General Director of UniqueICs Company. Scientific interests: systems of computer-aided design of integrated schemes and simulation methods; image processing and compression, pattern recognition. Aleksei Mikhailovich Fartukov. Born 1979. Graduated from the Moscow State Institute of Electronic Technology in 2001. Works as engineer-programmer inUniqueICs Company. Scientific interests: image processing and compression, pattern recognition.