基于稀疏表示的人脸姿态估计研究

针对人脸光照、遮挡、身份、表情等因素变化的人脸姿态估计难题,结合稀疏表示分类(SRC)方法的优秀识别性能,对SRC理论进行了深入分析,并将其应用于人脸姿态分类.为了解决姿态估计中人脸光照、噪声和遮挡变化问题,将人脸姿态离散化为不同的子空间,每个子空间对应一个类别,据此,提出基于字典学习与稀疏约束的人脸姿态识别方法.通过在公开的XJTU和PIE人脸库上实验表明:所研究的方法对人脸光照、噪声和遮挡变化具有鲁棒性.     

一种基于独立成分分析和最小最大概率机的人脸识别系统

提出了一种基于独立成分分析和最小最大概率机的人脸识别系统。该系统首先从摄像头中捕捉包含人脸的实时图像，利用haar特征人脸检测算法定位人脸区域，并将其从原始图像中分割出来。为了更好地提取有效特征，采用了ICA的特征提取方式，结合改进误差估计的最小最大概率机的分类方法对输入的测试图像进行识别。实验证明，该系统能够快速有效地处理实时状态下的人脸识别任务，准确率达到了96．8％，并且对多姿态的人脸具有一定的鲁棒性。     

复杂光照环境下视频序列人脸姿态跟踪方法研究

复杂光照环境下难以实现人脸姿态的高精准跟踪,为此研究复杂光照环境下视频序列人脸姿态跟踪方法,提升跟踪进度和效率。通过高频加强滤波和直方图均衡技术光照补偿视频序列图像,采用参考颜色表法匹配图像肤色序列特征,将匹配到的图像肤色序列特征作为待分类样本输入Boosting分类器,利用改进Adaboost算法自动挑选特征,检测待跟踪人脸姿态特征,以此为基础,通过创建二维肤色高斯模型,并不断更新肤色模型,克服复杂光照条件对肤色带来的影响,实现复杂光照环境下人脸姿态跟踪。实验结果表明,该方法对于平移、缩放、旋转、光照及遮挡等复杂光照环境下的人脸姿态都能较好跟踪,在摄像机固定状态下,跟踪准确率为98.57%,计算时间为1.5 s,在摄像机运动状态下,跟踪准确率为93.33%,计算时间为1.86 s,抗噪声干扰性能较为优越。     

基于PCA与SVM的头部姿态识别及应用

针对头部姿态识别在复杂背景和变化光照情况下准确率低的问题,提出了一种有效识别图像序列中头部姿态的方法.首先运用Adaboost算法提取出图像序列中不同姿态的人脸图像,通过主成分分析方法(PCA)提取人脸姿态特征;然后使用支持向量机构(SVM)造多分类器对提取的特征分类从而实现头部姿态识别;最后设计了五种不同的头部姿态在变化光照下与智能轮椅进行人机交互实验.实验结果表明该方法实时性高,抗光照变化性能强,识别率高达92.2％.     

基于生成对抗网络的多姿态人脸识别算法

头部姿态角转换会造成人脸成像多姿态变化,人脸离散数据的高斯分布混乱,无法准确地反映人脸多姿态的任意性和连续性,存在识别效果差的问题。引入生成对抗网络理论,设计多姿态人脸识别算法。对获取到的不同角度人脸图像,实施多姿态人脸校正与旋转残差注意力计算,解决当前头部姿态估计方法对不同人脸兴趣区域不稳健的问题。设计生成对抗网络进行双路循环优化,在生成的对抗网络中,参考CASIA-Net网络结构,使用深层次网络结构,每一层都有一个3*3的卷积核。所提出的设计可以降低网络参数,增强网络的非线性度,实现高效的面部特征提取,构建人脸多姿态识别模型,并完成人脸识别。通过实验结果表明,所提算法针对多姿态人脸识别效果好,在人脸不同姿态变化过程中,识别率始终在97%以上,更适用于多姿态人脸识别。     

一种基于动态贝叶斯网的视觉注意力识别方法

头部姿态估计是识别用户视觉注意力目标的主要依据.但在实际应用场合下,大范围头部姿态、低分辨率图像以及光照变化等因素使得可靠、准确的头部姿态估计难以实现.针对这些困难,提出一种基于动态贝叶斯网模型的视觉注意力目标识别方法.通过人脸图像与多个人脸姿态类别的相似度向量对头部姿态进行度量而不是显式的计算具体姿态值.模型融合多注意力目标、多用户位置、多摄像机图像等因素间的概率依赖关系并进行联合推理.智能厨房原型环境下的实验结果表明提出的模型是有效的.     

改进SRC算法在人脸识别中的应用

稀疏表示分类算法(Sparse Representation-based Classification,SRC)在人脸数据库上有很高的识别性能。然而,对于姿态变化,SRC的识别效果并不理想。针对SRC算法不能解决测试样本与训练样本存在偏移误差的问题,本文提出了基于SRC的改进算法。该算法将每一类的训练样本单独作为训练字典,利用迭代校正和基于金字塔分层机构的运动偏移估计方法得到最终的偏移量,最后对校正后的测试样本使用SRC算法实现分类。实验结果表明该方法对于有偏移误差的人脸图像具有较好的鲁棒性及识别率。     

基于形状匹配变形模型的三维人脸重构

为了从多幅人脸图像构造三维人脸结构,通常需要自动提取不同图像中的对应特征点,这往往是很难完成的.为了避免这个困难,本文建立了一个基于形状匹配的三维变形模型,在保证形状最佳匹配的条件下,实现对人脸图像姿态的估计和三维人脸重构.模型采用径向基函数对通用头部模型进行变形,用形状上下文来描述点之间的形状相似性,形状距离用来描述头部模型和人脸图像整体形状上的相似性,从而实现形状最佳匹配意义上的三维重构.实验表明,本文的算法只需要在人脸图像中提取特征点集,不需进行配准,就可以恢复出令人满意的三维头部结构.     

融合Mel系数和kNN的语音端点检测

语音端点检测是语音处理过程中的重要环节。为了提高在不同噪声环境下语音端点检测的准确率和鲁棒性,提出了融合语音Mel频率倒普系数和kNN分类算法相的语音端点检测方法。该方法提取语音信号的Mel频率倒普系数作为语音特征参数,然后将特征向量作为kNN分类的输入进行训练学习,建立语音端点检测模型,并进行仿真实验,结果表明该方法是一种准确率高、鲁棒性强的语音端点检测方法。     

结合Gabor特征和深度信念网络的人脸姿态分类

针对多姿态人脸图像分类存在的困难,提出了一种基于Gabor特征和深度信念网络(DBN)的近邻元分析(NCA)方法,通过提取Gabor多姿态人脸图像的尺度图并将其进行融合,从而对多姿态人脸图像具有较好的区分度,利用融合后的特征图来训练样本并作为深度信念网络的输入图像,结合NCA分析对训练样本进行线性变化以寻找到一个更有利于类别分类的线性子空间,提供足够大的数据集来估算模型参数进而对多姿态人脸图像进行分类.对ORL人脸数据集测试结果表明,多姿态人脸分类数据量为1616和2432之间时的平均分类正确率分别为86.67％、84.00％、90.67％和86.67％,与PCA、LDA和RCA三种算法相比,其分类准确率都得到了提高,实验结果验证了这种针对多姿态人脸图像的分类算法的有效性.     

Head pose estimation based on face symmetry analysis

This paper addresses the problem of head pose estimation in order to infer non-intrusive feedback from users about gaze attention. The proposed approach exploits the bilateral symmetry of the face. Size and orientation of the symmetrical area of the face is used to estimate roll and yaw poses by the mean of decision tree model. The approach does not need the location of interest points on face and presents robustness to partial occlusions. Tests were performed on different datasets (FacePix, CMU PIE, Boston University) and our approach coped with variability in illumination and expressions. Results demonstrate that the changes in the size of the regions that contain a bilateral symmetry provide accurate pose estimation.     

An image registration algorithm based on phase correlation and the classical Lucas–Kanade technique

Image registration is defined as an important process in image processing in order to align two or more images. A new image registration algorithm for translated and rotated pairs of 2D images is presented in order to achieve subpixel accuracy and spend a small fraction of computation time. To achieve the accurate rotation estimation, we propose a two-step method. The first step uses the Fourier Mellin Transform and phase correlation technique to get the large rotation, then the second one uses the Fourier Mellin Transform combined with an enhance Lucas–Kanade technique to estimate the accurate rotation. For the subpixel translation estimation, the proposed algorithm suggests an improved Hanning window as a preprocessing task to reduce the noise in images then achieves a subpixel registration in two steps. The first step uses the spatial domain approach which consists of locating the peak of the cross-correlation surface, while the second uses the frequency domain approach, based on low-frequency (aliasing-free part) of aliased images. Experimental results presented in this work show that the proposed algorithm reduces the computational complexities with a better accuracy compared to other subpixel registration algorithms.     

基于自适应非局部均值的CBCT投影数据去噪算法

为了准确地获取放疗摆位信息,并减低临床患者 所接收的辐射剂量,提出了一种基于非局部均值(NLM)的锥形束计算层析(CBCT)投影数据去 噪算法。首先,计算在不同投射角度下获取的CBCT投影数据的噪声标准差、边缘信息和纹理 子块的平均梯度值的均值,确 定与该角度投影数据相适应的滤波强度值;然后,采用改进的NLM算法对投影数 据进行去噪处理;最后,经过三维重 建获得较高质量的CBCT图像。还对基于子块分割的噪声估计算法进行了改进,使其更适 用于CBCT投影数据的噪声估 计。实验结果表明,本文算法能够有效估计投影数据的噪声水平,去噪效果优于其它几种算 法,在去除噪声的同时,还能很 好地保留图像的细节信息,并可增强图像的对比度,有利于准确获取摆位信息和医生的临 床诊断。     

一种Alpha稳定分布噪声下的体积目标长度估计方法

为了解决Alpha稳定分布噪声环境下运动舰船目标的长度估计问题,该文借鉴非线性变换抑制脉冲噪声以及多普勒目标运动特性估计思想,提出基于广义时频分析(G-TFA)和最小二乘估计的运动目标长度估计方法。该方法首先利用G-TFA获取Alpha稳定分布噪声环境下运动目标的多普勒频率,然后利用最小二乘方法估计出目标航速和不同位置的横正时刻,最后利用上述估计结果计算目标长度。以广义Winger-Ville分布(G-WVD)为例,从理论上推导了G-TFA在Alpha稳定分布噪声环境下具有提取目标多普勒特征的能力,并通过仿真实验验证了该算法在中低混合信噪比下的稳健性。与现有算法相比,该文所提算法不需要估计噪声特征指数,算法性能优于基于传统时频分析的估计方法。     

基于边缘剔除的陆基高光谱图像噪声评估方法

为了准确地对陆基成像条件下高光谱图像的噪声水平进行评估, 提出了一种基于边缘剔除后残差调整的局部标准差法。首先将获取的高光谱图像分成若干个大小合适的子块, 而后利用Canny边缘检测算子检测出图像的边缘信息, 判断并剔除其中含有边缘的子块, 将剔除边缘子块后的均匀子块采用多元线性回归后求取残差的方法进行噪声估计。结果表明, 对同一幅陆基高光谱图像的不同子区域进行4×4像元与8×8像元分块, 得到的噪声总误差值分别为1.985×103与2.197×103。该噪声估计方法对陆基成像条件下高光谱图像的噪声评估具有较强的鲁棒性, 可为后续陆基高光谱图像处理与应用提供参考。     

Speckle noise removal based on structural convolutional neural networks with feature fusion for medical image

Application of convolutional neural networks (CNNs) for image additive white Gaussian noise (AWGN) removal has attracted considerable attentions with the rapid development of deep learning in recent years. However, the work of image multiplicative speckle noise removal is rarely done. Moreover, most of the existing speckle noise removal algorithms are based on traditional methods with human priori knowledge, which means that the parameters of the algorithms need to be set manually. Nowadays, deep learning methods show clear advantages on image feature extraction. Multiplicative speckle noise is very common in real life images, especially in medical images. In this paper, a novel neural network structure is proposed to recover noisy images with speckle noise. Our proposed method mainly consists of three subnetworks. One network is rough clean image estimate subnetwork. Another is subnetwork of noise estimation. The last one is an information fusion network based on U-Net and several convolutional layers. Different from the existing speckle denoising model based on the statistics of images, the proposed network model can handle speckle denoising of different noise levels with an end-to-end trainable model. Extensive experimental results on several test datasets clearly demonstrate the superior performance of our proposed network over state-of-the-arts in terms of quantitative metrics and visual quality.     

Model-based temporal object verification using video

An approach to model-based dynamic object verification and identification using video is proposed. From image sequences containing the moving object, we compute its motion trajectory. Then we estimate its three-dimensional (3-D) pose at each time step. Pose estimation is formulated as a search problem, with the search space constrained by the motion trajectory information of the moving object and assumptions about the scene structure. A generalized Hausdorff (1962) metric, which is more robust to noise and allows a confidence interpretation, is suggested for the matching procedure used for pose estimation as well as the identification and verification problem. The pose evolution curves are used to assist in the acceptance or rejection of an object hypothesis. The models are acquired from real image sequences of the objects. Edge maps are extracted and used for matching. Results are presented for both infrared and optical sequences containing moving objects involved in complex motions     

光束质量稳健估计非线性拟合求解方法

激光光束质量M2因子在测量过程中,数据不可避免的存在随机和非随机两类误差,为了减小测量过程中受到误差的影响以提高系统的测量精度和稳定性,针对不同误差影响进行分析并提出相应的解决方法。在随机误差情况下,对传统双曲线拟合的正规方程组解法进行了改进并提出了加权拟合的正规方程组解法;而对于非随机误差情况,双曲线拟合尚未发现有效的解决方案,为此提出了基于稳健估计的双曲线拟合方法,并对此进行了深入的理论研究。实验表明,在数据存在非随机的情况下采用稳健估计的方法,可以克服非随机误差对参数估值产生的影响,使其拟合优度接近最优值,稳健估计的误差要比正规方程组的误差低一个数量级。因此,稳健估计可有效地提高测量M2因子的测量精度,对评价光束质量方面有着重要的应用价值。     

An adaptive semiparametric and context-based approach tounsupervised change detection in multitemporal remote-sensingimages

A novel automatic approach to the unsupervised identification of changes in multitemporal remote-sensing images is proposed. This approach, unlike classical ones, is based on the formulation of the unsupervised change-detection problem in terms of the Bayesian decision theory. In this context, an adaptive semiparametric technique for the unsupervised estimation of the statistical terms associated with the gray levels of changed and unchanged pixels in a difference image is presented. Such a technique exploits the effectivenesses of two theoretically well-founded estimation procedures: the reduced Parzen estimate (RPE) procedure and the expectation-maximization (EM) algorithm. Then, thanks to the resulting estimates and to a Markov random field (MRF) approach used to model the spatial-contextual information contained in the multitemporal images considered, a change detection map is generated. The adaptive semiparametric nature of the proposed technique allows its application to different kinds of remote-sensing images. Experimental results, obtained on two sets of multitemporal remote-sensing images acquired by two different sensors, confirm the validity of the proposed approach.