Toward improved ranking metrics

In many computer vision algorithms, a metric or similarity measure is used to determine the distance between two features. The Euclidean or SSD (sum of the squared differences) metric is prevalent and justified from a maximum likelihood perspective when the additive noise distribution is Gaussian. Based on real noise distributions measured from international test sets, we have found that the Gaussian noise distribution assumption is often invalid. This implies that other metrics, which have distributions closer to the real noise distribution, should be used. In this paper, we consider three different applications: content-based retrieval in image databases, stereo matching, and motion tracking. In each of them, we experiment with different modeling functions for the noise distribution and compute the accuracy of the methods using the corresponding distance measures. In our experiments, we compared the SSD metric, the SAD (sum of the absolute differences) metric, the Cauchy metric, and the Kullback relative information. For several algorithms from the research literature which used the SSD or SAD, we showed that greater accuracy could be obtained by using the Cauchy metric instead.     

基于模糊不确定性的自适应采样

提出一种用于MonteCarlo全局光照的自适应采样方法,使得合成图像时对每个像素采用不同的采样数量,以提高间接光照的表现效果,降低图像总体噪声水平．考虑到图像或像素噪声水平的评价具有内在的模糊不确定性,基于模糊理论,以像素样本光照为基本元素建立模糊集合,利用模糊集的模糊度提出一种新的像素噪声水平评价标准．在新评价标准的基础之上实施自适应采样,首先对像素进行少量采样,然后根据新标准评价其噪声水平,并有针对性地对噪声水平较高的像素使用较多的采样样本．通过大量实验,验证了文中方法比已有的自适应采样方法更好．     

Quantum deconvolution

We propose a method for stably removing noise from measurements of a quantum many-body system. The question is cast to a linear inverse problem by using a quantum Fischer information metric as figure of merit. This requires the ability to compute the adjoint of the noise channel with respect to the metric, which can be done analytically when the metric is evaluated at a Gaussian (quasi-free) state. This approach can be applied effectively to n-point functions of a quantum field theory. For translation invariant noise, this yields a stable deconvolution method on the first moments of the field which differs from what one would obtain from a purely classical analysis.     

基于最大相关熵准则的鲁棒度量学习算法

度量亦称距离函数,是度量空间中满足特定条件的特殊函数,一般用来反映数据间存在的一些重要距离关系.而距离对于各种分类聚类问题影响很大,因此度量学习对于这类机器学习问题有重要影响.受到现实存在的各种噪声影响,已有的各种度量学习算法在处理各种分类问题时,往往出现分类准确率较低以及分类准确率波动大的问题.针对该问题,本文提出一种基于最大相关熵准则的鲁棒度量学习算法.最大相关熵准则的核心在于高斯核函数,本文将其引入到度量学习中,通过构建以高斯核函数为核心的损失函数,利用梯度下降法进行优化,反复测试调整参数,最后得到输出的度量矩阵.通过这样的方法学习到的度量矩阵将有更好的鲁棒性,在处理受噪声影响的各种分类问题时,将有效地提高分类准确率.本文将在一些常用机器学习数据集（UCI）还有人脸数据集上进行验证实验.     

Exponential stability of discrete-time filters for bounded observation noise

This paper proves exponential asymptotic stability of discrete-time filters for the estimation of solutions to stochastic difference equations, when the observation noise is bounded. No assumption is made on the ergodicity of the signal. The proof uses the Hilbert projective metric, introduced into filter stability analysis by Atar and Zeitouni [1,2]. It is shown that when the signal noise is sufficiently regular, boundedness of the observation noise implies that the filter update operation is, on average, a strict contraction with respect to the Hilbert metric. Asymptotic stability then follows.     

基于图像质量评价的自动图像复原技术

针对很难使用一种图像复原技术来满足实际的成像系统应用需求的情况,提出了一种基于无参考图像质量评价的自动图像复原技术。该技术在图像复原中引入图像质量评价,使图像复原技术变为主动图像复原。为应对图像降质过程的复杂性,介绍了基于图像质量评价的多模式图像复原框架。为适应由于图像复原过程的病态性导致的图像误差问题,提出了一种新的无参考图像质量评价方法,该算法通过在图像质量算子中引入图像相似矩阵和图像复原趋势矩阵,使其能适应复原对于图像结构或噪声结构变化。实验表明该图像质量评价算子的有效性,可以满足自动图像复原对图像质量评价算法的要求。     

A belief function distance metric for orderable sets

This paper describes a new metric for characterizing conflict between belief assignments. The new metric, specifically designed to quantify conflict on orderable sets, uses a Hausdorff-based measure to account for the distance between focal elements. This results in a distance metric that can accurately measure conflict between belief assignments without saturating simply because two assignments do not have common focal elements. The proposed metric is particularly attractive in sensor fusion applications in which belief is distributed on a continuous measurement space. Several example cases demonstrate the proposed metric’s performance, and comparisons with other common measures of conflict show the significant benefit of using the proposed metric in cases where a sensor’s error and noise characteristics are not known precisely a priori.     

No-Reference Image Quality Assessment for Image Auto-Denoising

This paper proposes two new non-reference image quality metrics that can be adopted by the state-of-the-art image/video denoising algorithms for auto-denoising. The first metric is proposed based on the assumption that the noise should be independent of the original image. A direct measurement of this dependence is, however, impractical due to the relatively low accuracy of existing denoising method. The proposed metric thus tackles the homogeneous regions and highly-structured regions separately. Nevertheless, this metric is only stable when the noise level is relatively low. Most denoising algorithms reduce noise by (weighted) averaging repeated noisy measurements. As a result, another metric is proposed for high-level noise based on the fact that more noisy measurements will be required when the noise level increases. The number of measurements before converging is thus related to the quality of noisy images. Our patch-matching based metric proposes to iteratively find and add noisy image measurements for averaging until there is no visible difference between two successively averaged images. Both metrics are evaluated on LIVE2 (Sheikh et al. in LIVE image quality assessment database release 2: 2013) and TID2013 (Ponomarenko et al. in Color image database tid2013: Peculiarities and preliminary results: 2005) data sets using standard Spearman and Kendall rank-order correlation coefficients (ROCC), showing that they subjectively outperforms current state-of-the-art no-reference metrics. Quantitative evaluation w.r.t. different level of synthetic noisy images also demonstrates consistently higher performance over state-of-the-art non-reference metrics when used for image denoising.     

基于体半径函数的网格分割算法

为解决现有网格分割算法对表面局部特征和噪声敏感的问题,定义一种刻画模型体属性的测度——体半径,该测度具有姿态不变性和噪声鲁棒性.在此基础上,提出一种网格层次分割算法.计算模型顶点的体半径值,采用二状态高斯混合模型对体半径分布拟合,进行连通分量标注,获取模型的各子部件,结合图切分技术得到有视觉意义的分割结果.实验结果表明...     

非贪婪的鲁棒性度量学习算法

目的 度量学习是机器学习与图像处理中依赖于任务的基础研究问题。由于实际应用背景复杂,在大量不可避免的噪声环境下,度量学习方法的性能受到一定影响。为了降低噪声影响,现有方法常用L1距离取代L2距离,这种方式可以同时减小相似样本和不相似样本的损失尺度,却忽略了噪声对类内和类间样本的不同影响。为此,本文提出了一种非贪婪的鲁棒性度量学习算法——基于L2/L1损失的边缘费歇尔分析（marginal Fisher analysis based on L2/L1 loss,MFA-L2/L1）,采用更具判别性的损失,可提升噪声环境下的识别性能。方法 在边缘费歇尔分析（marginal Fisher analysis,MFA）方法的基础上,所提模型采用L2距离刻画相似样本损失、L1距离刻画不相似样本损失,同时加大对两类样本的惩罚程度以提升方法的判别性。首先,针对模型非凸带来的求解困难,将目标函数转为迭代两个凸函数之差便于求解;然后,受DCA（difference of convex functions algorithm）思想启发,推导出非贪婪的迭代求解算法,求得最终度量矩阵;最后,算法的理论证明保证了迭代算法的收敛性。结果 在5个UCI（University of California Irrine）数据集和7个人脸数据集上进行对比实验：1）在不同程度噪声的5个UCI数据集上,MFA-L2/L1算法最优,且具有较好的抗噪性,尤其在30%噪声程度的Seeds和Wine数据集上,与次优方法LDA-NgL1（non-greedy L1-norm linear discriminant analysis））相比,MFA-L2/L1的准确率高出9%;2）在不同维度的AR和FEI人脸数据集上的实验,验证了模型采用L1损失、采用L2损失提升了模型的判别性;3）在Senthil、Yale、ORL、Caltech和UMIST人脸数据集的仿真实验中,MFA-L2/L1算法呈现出较强鲁棒性,性能排名第1。结论 本文提出了一种基于L2/L1损失的鲁棒性度量学习模型,并推导了一种便捷有效的非贪婪式求解算法,进行了算法收敛性的理论分析。在不同数据集的不同噪声情况下的实验结果表明,所提算法具有较好的识别率和鲁棒性。     

Novel fuzzy active contour model with kernel metric for image segmentation

In this paper, a novel region-based fuzzy active contour model with kernel metric is proposed for a robust and stable image segmentation. This model can detect the boundaries precisely and work well with images in the presence of noise, outliers and low contrast. It segments an image into two regions – the object and the background by the minimization of a predefined energy function. Due to the kernel metric incorporated in the energy and the fuzziness of the energy, the active contour evolves very stably without the reinitialization for the level set function during the evolution. Here the fuzziness provides the model with a strong ability to reject local minima and the kernel metric is employed to construct a nonlinear version of energy function based on a level set framework. This new fuzzy and nonlinear version of energy function makes the updating of region centers more robust against the noise and outliers in an image. Theoretical analysis and experimental results show that the proposed model achieves a much better balance between accuracy and efficiency compared with other active contour models.     

Gradient entropy metric and p-Laplace diffusion constraint-based algorithm for noisy multispectral image fusion

Noise is easily mistaken as useful features of input images, and therefore, significantly reducing image fusion quality. In this paper, we propose a novel gradient entropy metric and p-Laplace diffusion constraint-based method. Specifically, the method is based on the matrix of structure tensor to fuse the gradient information. To minimize the negative effects of noise on the selections of image features, the gradient entropy metric is proposed to construct the weight for each gradient of input images. Particularly, the local adaptive p-Laplace diffusion constraint is constructed to further suppress noise when rebuilding the fused image from the fused gradient field. Experimental results show that the proposed method effectively preserves edge detail features of multispectral images while suppressing noise, achieving an optimal visual effect and more comprehensive quantitative assessments compared to other existing methods.     

Spectral difference for statistical model-based speech enhancement in speech recognition

In this paper, we propose a statistical model-based speech enhancement technique using the spectral difference scheme for the speech recognition in virtual reality. In the analyzing step, two principal parameters, the weighting parameter in the decision-directed (DD) method and the long-term smoothing parameter in noise estimation, are uniquely determined as optimal operating points according to the spectral difference under various noise conditions. These optimal operating points, which are specific according to different spectral differences, are estimated based on the composite measure, which is a relevant criterion in terms of speech quality. An efficient mapping function is also presented to provide an index of the metric table associated with the spectral difference so that operating points can be determined according to various noise conditions for an on-line step. In the on-line speech enhancement step, different parameters are chosen on a frame-by-frame basis under the metric table of the spectral difference. The performance of the proposed method is evaluated using objective and subjective speech quality measures in various noise environments. Our experimental results show that the proposed algorithm yields better performances than conventional algorithms.     

Sobolev 广义度量下的各向异性扩散模型

给出了Sobolev空间的一种广义度量, 在该度量下提出了一个新的各向异性增强、扩散方程. 广义度量中的变系数, 较好地控制了方程的扩散行为, 使得新模型不仅能有效增强图像的细节特征, 而且能在噪声去除和边缘保护之间取得较好的平衡, 同时给出了相应的隐式离散算法. 仿真实验结果表明, 新模型和算法是行之有效的.     

基于RefineNet的端到端语音增强方法

为提高神经网络对语音信号时域波形的直接处理能力,提出了一种基于RefineNet的端到端语音增强方法.本文构建了一个时频分析神经网络,模拟语音信号处理中的短时傅里叶变换,利用RefineNet网络学习含噪语音到纯净语音的特征映射.在模型训练阶段,用多目标联合优化的训练策略将语音增强的评价指标短时客观可懂度(Short-time objective intelligibility,STOI)与信源失真比(Source to distortion ratio,SDR)融入到训练的损失函数.在与具有代表性的传统方法和端到端的深度学习方法的对比实验中,本文提出的算法在客观评价指标上均取得了最好的增强效果,并且在未知噪声和低信噪比条件下表现出更好的抗噪性.     

基于RefineNet的端到端语音增强方法

为提高神经网络对语音信号时域波形的直接处理能力,提出了一种基于RefineNet的端到端语音增强方法.本文构建了一个时频分析神经网络,模拟语音信号处理中的短时傅里叶变换,利用RefineNet网络学习含噪语音到纯净语音的特征映射.在模型训练阶段,用多目标联合优化的训练策略将语音增强的评价指标短时客观可懂度(Short-time objective intelligibility,STOI)与信源失真比(Source to distortion ratio,SDR)融入到训练的损失函数.在与具有代表性的传统方法和端到端的深度学习方法的对比实验中,本文提出的算法在客观评价指标上均取得了最好的增强效果,并且在未知噪声和低信噪比条件下表现出更好的抗噪性.     

Anisotropic noise samples

We present a practical approach to generate stochastic anisotropic samples with Poisson-disk characteristic over a two-dimensional domain. In contrast to isotropic samples, we understand anisotropic samples as non-overlapping ellipses whose size and density match a given anisotropic metric. Anisotropic noise samples are useful for many visualization and graphics applications. The spot samples can be used as input for texture generation, e.g., line integral convolution (LIC), but can also be used directly for visualization. The definition of the spot samples using a metric tensor makes them especially suitable for the visualization of tensor fields that can be translated into a metric. Our work combines ideas from sampling theory and mesh generation. To generate these samples with the desired properties we construct a first set of non-overlapping ellipses whose distribution closely matches the underlying metric. This set of samples is used as input for a generalized anisotropic Lloyd relaxation to distribute noise samples more evenly. Instead of computing the Voronoi tessellation explicitly, we introduce a discrete approach which combines the Voronoi cell and centroid computation in one step. Our method supports automatic packing of the elliptical samples, resulting in textures similar to those generated by anisotropic reaction-diffusion methods. We use Fourier analysis tools for quality measurement of uniformly distributed samples. The resulting samples have nice sampling properties, for example, they satisfy a blue noise property where low frequencies in the power spectrum are reduced to a minimum.     

利用无参考图像内容量度优化非局部均值去噪方法的参数

目的 非局部均值NLM（non-local means）方法是一种有效的数字图像去噪方法。然而,在实际去噪过程中,非局部均值的衰减参数通常是固定的而且无法随着图像的变化而作适应性的调整。为了使非局部均值方法更加有效, 提出一种将适用于多种噪声分布的无参考图像内容量度（表示为Q）引入NLM的迭代方法,来优化固定的衰减参数。方法 首先,针对普通图像的去噪,利用量度Q来测量每一次调整衰减参数后所对应的去噪结果的图像质量,凭借该迭代机制找到Q的最大值,从而获得最优的图像去噪结果;其次,将该量度用于MRI（magnetic resonance imaging）图像的去噪,利用Q来度量图像所含结构信息（如纹理和边缘）,进而调整用于MRI图像去噪的无偏非局部均值法的衰减参数。结果 实验结果显示,本文方法提升了去噪结果的峰值信噪比（PSNR）,并且本文方法的去噪结果在视觉上看起来比用传统方法得到的结果更清晰。结论 利用无参考图像内容量度Q来优化NLM方法的衰减参数,使得NLM方法能够针对不同的图像自适应地调整衰减参数以取得最优的去噪效果。实验结果表明用图像内容量度Q来优化非局部均值法的参数是有效的。     

Time complexity and model complexity of fast identification ofcontinuous-time LTI systems

The problem of fast identification of continuous-time systems is formulated in the metric complexity theory setting. It is shown that the two key steps to achieving fast identification, i.e., optimal input design and optimal model selection, can be carried out independently when the true system belongs to a general a priori set. These two optimization problems can be reduced to standard Gel'fand and Kolmogorov n-width problems in metric complexity theory. It is shown that although arbitrarily accurate identification can be achieved on a small time interval by reducing the noise-signal ratio and designing the input carefully, identification speed is limited by the metric complexity of the a priori uncertainty set when the noise/signal ratio is fixed     

Information quality model and optimization for 802.15.4-based wireless sensor networks

High information quality is a paramount requirement for wireless sensor network (WSN) monitoring applications. However, it is challenging to achieve a cost effective information quality solution due to unpredictable environment noise and events, unreliable wireless channel and network bandwidth, and sensor resource and energy constraints. Specifically, the dynamic and unreliable nature of WSNs make it difficult to pre-determine optimum sensor rates and predict packet loss. To address this problem, we present an information quality metric which characterizes information quality based on the sampling frequency of sensor nodes and the packet loss rate during network transmission. Our fundamental quality metric is based on signal-to-noise ratio and is therefore application independent. Based on our metric, a quality-aware scheduling system (QSS) is developed, which exploits cross-layer control of sensor nodes to effectively schedule data sensing and forwarding. Particularly, we develop and evaluate several QSS scheduling mechanisms: passive, reactive and perceptive. These mechanisms can adapt to environment noise, bandwidth variation and wireless channel collisions by dynamically controlling sensor rates and phase. Our experimental results indicate that our QSS is a novel and effective approach to improve information quality for WSNs.