Monte Carlo despeckling of transrectal ultrasound images of the prostate

In this paper, a novel stochastic method is developed for despeckling transrectal ultrasound (TRUS) images of the prostate. By incorporating the circular probe acquisition particularities and speckle noise statistics of TRUS images of the prostate into a likelihood-weighted Monte Carlo estimation scheme, the proposed method can better remove speckle noise while preserving image structures and details that are relevant for image screening, allowing for a better delineation of the lesion contour. Our in silico and in vivo experimental results are promising, which was confirmed by a clinical evaluation of the in vivo test cases by experienced clinicians, and indicate that our method potentially can perform better than other previously proposed methods.     

Unsupervised Image Reconstruction for Gradient-Domain Volumetric Rendering

Gradient-domain rendering can highly improve the convergence of light transport simulation using the smoothness in image space. These methods generate image gradients and solve an image reconstruction problem with rendered image and the gradient images. Recently, a previous work proposed a gradient-domain volumetric photon density estimation for homogeneous participating media. However, the image reconstruction relies on traditional L1 reconstruction, which leads to obvious artifacts when only a few rendering passes are performed. Deep learning based reconstruction methods have been exploited for surface rendering, but they are not suitable for volume density estimation. In this paper, we propose an unsupervised neural network for image reconstruction of gradient-domain volumetric photon density estimation, more specifically for volumetric photon mapping, using a variant of GradNet with an encoded shift connection and a separated auxiliary feature branch, which includes volume based auxiliary features such as transmittance and photon density. Our network smooths the images on global scale and preserves the high frequency details on a small scale. We demonstrate that our network produces a higher quality result, compared to previous work. Although we only considered volumetric photon mapping, it's straightforward to extend our method for other forms, like beam radiance estimation.     

基于多尺度分析和分形的人群密度估计方法

人群的密度估计对于建筑物的设计和人群安全至关重要。论文提出了一种新的人群密度自动估计方法。对于低密度人群图像采用基于像素统计的方法,对于较高密度人群图像采用基于多尺度分析和分形的纹理分析方法,并应用支撑向量机进行人群密度等级分类。对人群图像集的实验表明,该方法较以前的方法更为准确有效。     

基于完全局部二值模式的人群密度估计

为了实时监控重要场所的人群密度、采取有效措施疏散高密度人群,避免人群密度过大而引发事故,造成生命和财产的损失,提出了一种基于完全局部二值模式的人群密度估计方法.该方法提取人群图像的3种局部纹理特征,建立了3-D联合直方图统计模型,用卡方距离最近邻方法对人群密度级别进行分类,实现了特定场景下人群密度的监测.对比实验结果表明了该方法能兼顾任意密度级别人群图像的分类,不仅准确率高,而且实时性好,同时对场景背景具有较强的鲁棒性.     

生成对抗网络及其在图像生成中的应用研究综述

生成对抗网络(GAN)是无监督学习领域最近几年快速发展的一个研究方向,其主要特点是能够以一种间接的方式对一个未知分布进行建模.在计算机视觉研究领域中,生成对抗网络有着广泛的应用,特别是在图像生成方面,与其他的生成模型相比,生成对抗网络不仅可以避免复杂的计算,而且生成的图像质量也更好.因此,本文将对生成对抗网络及其在图像生成中的研究进展做一个小结和分析:本文首先从模型的架构、目标函数的设计、生成对抗网络在训练中存在的问题、以及如何处理模式崩溃问题等角度对生成对抗网络进行一个详细地总结和归纳;其次介绍生成对抗网络在图像生成中的两种方法;随后对一些典型的、用来评估生成图像质量和多样性的方法进行小结,并且对基于图像生成的应用进行详细分析;最后对生成对抗网络和图像生成进行总结,同时对其发展趋势进行一个展望.     

基于流形学习的彩色遥感图像分维数估算

目的 纹理特征提取一直是遥感图像分析领域研究的热点和难点。现有的纹理特征提取方法主要集中于研究单波段灰色遥感图像,如何提取多波段彩色遥感图像的纹理特征,是多光谱遥感的研究前沿。方法 提出了一种基于流形学习的彩色遥感图像分维数估算方法。该方法利用局部线性嵌入方法,对由颜色属性所组成的5-D欧氏超曲面进行维数简约处理;再将维数简约处理后的颜色属性用于分维数估算。结果 利用Landsat-7遥感卫星数据和GeoEye-1遥感卫星数据进行实验,结果表明,同Peleg法和Sarkar法等其他分维数估算方法相比,本文方法具有较小的拟合误差。其中,其他4种对比方法所获拟合误差E平均值分别是本文方法所获得拟合误差E平均值的26.2倍、5倍、26.3倍、5倍。此外,本文方法不仅可提供具有较好分类特性的分维数,而且还能提供相对于其他4种对比方法更加稳健的分维数。结论 在针对中低分辨率的真彩遥感图像和假彩遥感图像以及高分辨率彩色合成遥感图像方面,本文方法能够利用不同地物所具有颜色属性信息,提取出各类型地物所对应的纹理信息,有效地改善了分维数对不同地物的区分能力。这对后续研究各区域中不同类型地物的分布情况及针对不同类型地物分布特点而制定区域规划及开发具有积极意义。     

基于近似密度函数的医学图像聚类分析研究

针对医学图像数据难以用数学模型来表述和聚类的问题，提出一种基于近似密度函数的医学图像聚类分析方法．该方法采用核密度估计模型来构造近似密度函数，利用爬山策略来提取聚类模式．基于真实的人体腹部医学图像数据集的实验结果表明，该方法可以取得较好的聚类效果．     

Image super-resolution by dictionary concatenation and sparse representation with approximate L0 norm minimization

This paper proposes a different image super-resolution (SR) reconstruction scheme, based on the newly advanced results of sparse representation and the recently presented SR methods via this model. Firstly, we online learn a subsidiary dictionary with the degradation estimation of the given low-resolution image, and concatenate it with main one offline learned from many natural images with high quality. This strategy can strengthen the expressive ability of dictionary atoms. Secondly, the conventional matching pursuit algorithms commonly use a fixed sparsity threshold for sparse decomposition of all image patches, which is not optimal and even introduces annoying artifacts. Alternatively, we employ the approximate L₀ norm minimization to decompose accurately the patch over its dictionary. Thus the coefficients of representation with variant number of nonzero items can exactly weight atoms for those complicated local structures of image. Experimental results show that the proposed method produces high-resolution images that are competitive or superior in quality to results generated by similar techniques.     

基于核密度估计的图像自动标注方法

提出一种基于概率模型的图像自动语义标注方法,将图片自动标注看作一个多类分类问题,通过无参数的核密度估计,实现用含有共同标注词的图片组估计视觉特征和相应标注词之间关系的机制。选取表达能力较好的基于CPAM的视觉特征,无须对图像进行语义分割处理,有效提高核密度估计的效率。在基准数据集上进行实验,结果表明,该模型能够获得比当前其他相关方法更好的标注性能。     

Estimation of Edge Parameters and Image Blur Using Polynomial Transforms

A method is presented for detecting blurred edges in images and for estimating the following edge parameters: position, orientation, amplitude, mean value, and edge slope. The method is based on a local image decomposition technique called a polynomial transform. The information that is made explicit by the polynomial transform is well suited to detect image features, such as edges, and to estimate feature parameters. By using the relationship between the polynomial coefficients of a blurred feature and those of the a priori assumed (unblurred) feature in the scene, the parameters of the blurred feature can be estimated. The performance of the proposed edge parameter estimation method in the presence of image noise has been analyzed. An algorithm is presented for estimating the spread of a position-invariant Gaussian blurring kernel, using estimates at different edge locations over the image. First a single-scale algorithm is developed in which one polynomial transform is used. A critical parameter of the single-scale algorithm is the window size, which has to be chosen a priori. Since the reliability of the estimate for the spread of the blurring kernel depends on the ratio of this spread to the window size, it is difficult to choose a window of appropriate size a priori. The problem is overcome by a multiscale blur estimation algorithm where several polynomial transforms at different scales are applied, and the appropriate scale for analysis is chosen a posteriori. By applying the blur estimation algorithm to natural and synthetic images with different amounts of blur and noise, it is shown that the algorithm gives reliable estimates for the spread of the blurring kernel even at low signal-to-noise ratios.     

多模态硬币图像单应性矩阵预测

对不同成像条件下拍摄的硬币图像进行配准是硬币表面缺陷算法的前置任务。然而,基于互信息的传统多模态配准方法速度慢、精度低,现有的通过基于深度学习的单应性矩阵预测方法实现的图像配准只适用于单模态的任务。为此,提出一种基于深度学习的适用于多模态硬币图像的单应性矩阵预测方法,进而使用预测的单应性矩阵完成图像配准。首先,使用单应性矩阵预测层预测输入图像对间的单应性矩阵,使用单应性矩阵对待配准图像进行透视变换;然后,使用图像转换层将透视变换后的待配准图像和目标图像映射到同一域,图像转换层可在推理时去除从而减少推理时间;最后,计算同一域下的图像间的损失,并进行训练。实验表明,该方法在测试集上的平均距离误差为 3.417 像素,相较于基于互信息的传统多模态配准方法 5.575 像素的平均距离误差降低 38.71%。且配准单对图像耗时为 17.74 ms,远低于基于互信息的传统多模态配准方法的 6 368.49 ms。     

基于密度分类及组合特征的人数估计算法*

为了克服不同人群密度及所采用特征对人数估计的影响,提出了一种基于人群密度分类及组合特征的人数统计算法。该算法包括离线特征组合选取和在线实时估计两个阶段。在离线阶段,选取密度阈值将图像样本分为高、低密度两类,然后通过实验方法选取最优的特征组合。在线估计阶段首先通过分类器将样本分为高、低密度两类,然后利用离线阶段选取的特征组合训练得到高斯模型,并分别对两类样本进行人数估计。实验结果表明,与不分高低密度相比,平均估计误差由10.6%降至8.1%;与目前主流的人数估计算法相比,本文算法的平均估计误差也更小。     

亮度—梯度联合约束的车牌图像超分辨率重建

目的 受成像距离、光照条件、动态模糊等因素影响,监控系统拍摄的车牌图像往往并不具备较高的可辨识度。为改善成像质量,提升对车牌的识别能力,提出一种基于亮度与梯度联合约束的车牌图像超分辨率重建方法。方法 首先充分结合亮度约束和梯度约束的优势,实现对运动位移和模糊函数的精确估计;为抑制重建图像中的噪声与伪影,基于车牌图像的文字化特征,进一步确定了亮度与梯度联合约束的图像先验模型。结果 为验证该方法的有效性,利用监控系统获得4组车牌图像,分别进行模拟和真实的超分辨率重建实验。在模拟实验中将联合约束图像先验重建结果与拉普拉斯、Huber-Markov（HMRF）以及总变分（TV）先验的处理结果进行对比,联合约束先验对车牌纹理信息的恢复效果优于其他3种常见图像先验;同时,在模拟和真实实验中,将本文算法与双三次插值、传统最大后验概率、非线性扩散正则化和自适应范数正则化方法的超分辨率重建结果进行比较,模拟实验的结果表明,在不添加噪声情况下,该算法峰值信噪比（PSNR）和结构相似性（SSIM）指标分别为35.326 dB和0.958,优于其他4种算法;该算法在真实实验中,能够有效增强车牌图像纹理信息,获得较优的视觉效果,通过对重建车牌图像的字符识别精度比较,本文算法重建结果的识别精度远高于其他3种算法,平均字符差距为1.3。结论 模拟和真实图像序列的实验结果证明,基于亮度—梯度联合约束的超分辨率重建方法,能够降低运动和模糊等参数的估计误差,有效减少图像中存在的模糊和噪声,提高车牌的识别精度。该算法广泛适用于因光照变化、相对运动等因素影响下的低质量车牌图像超分辨率重建。     

Face posture estimation using eigen analysis on an IBR (image based rendered) database

In this paper, we present a novel representation of the human face for estimating the orientation of the human head in a two dimensional intensity image. The method combines the use of the much familiar eigenvalue based dissimilarity measure with image based rendering. There are two main components of the algorithm described here: the offline hierarchical image database generation and organization, and the online pose estimation stage. The synthetic images of the subject's face are automatically generated offline, for a large set of pose parameter values, using an affine coordinate based image reprojection technique. The resulting database is formally called as the IBR (or image based rendered) database. This is followed by the hierarchical organization of the database, which is driven by the eigenvalue based dissimilarity measure between any two synthetic image pair. This hierarchically organized database is a detailed, yet structured, representation of the subject's face. During the pose estimation of a subject in an image, the eigenvalue based measure is invoked again to search the synthetic (IBR) image closest to the real image. This approach provides a relatively easy first step to narrow down the search space for complex feature detection and tracking algorithms in potential applications like virtual reality and video-teleconferencing applications.     

编码曝光图像的L0正则化去模糊重建方法

目标的运动会导致其成像模糊。为了从模糊的图像中恢复清晰的目标图像,本文采用了编码曝光成像技术。与传统相机成像中快门一直处于开启的状态不同,编码曝光相机成像是在快门开启和闭合转换过程中成像。由于在时域快速转换的编码等效为频域较宽的滤波器,因此编码曝光成像有效地保留了目标的高频信息。为了从编码曝光图像中清晰地复原图像,本文设计了能保留图像高频细节的L ₀正则项约束的图像重建和模糊核估计方法。通过待重建图像与模糊核的交替迭代更新来完成图像重建。仿真合成图像和实际采集图像的实验表明,本方法对多种运动产生的模糊均有良好的图像复原效果。     

运动模糊图像PSF参数估计方法改进及图像复原

运动模糊图像复原的目的是改善运动图像质量,从而为图像处理任务提供高质量的清晰图像以保证算法能够准确获取图像信息,其中运动模糊图像的点扩散函数（PSF）求解是影响复原图像质量的关键步骤。针对现有运动模糊图像PSF参数估计方法中存在的估计误差大、有效估计范围有限等问题,在分析频谱图像特征的基础上,提出一种改进的PSF参数估计方法。通过图像增强处理和形态学变换去除频谱图像中的十字亮线和噪点干扰,获取形态合适的条纹图像以完成Radon变换检测。利用二值频谱图像的条纹特征自适应地控制形态学运算精度,从而保证算法的执行效率和鲁棒性。对条纹进行边缘测定,消除由条纹自身宽度导致的角度估计误差,以提高参数估计结果的精度。实验结果表明,该方法能够提高模糊参数估计的准确率和有效估计范围,由此构建的PSF能复原出更加清晰的重建图像,复原图像总体峰值信噪比不低于25 dB。     

多视点视差估计中基于图像块特征矢量和相似尺度的参考视点选择方法

在多视点视频处理中，采用多参考视点的视差估计方法能够在很大程度上消除各个视点间的冗余性，达到对多视点视频数据有效编码压缩的目的。研究了多视点信号处理中视差估计的多参考视点选择问题，采用快速Walsh变换谱构造图像块特征矢量，并提出了基于图像块特征和相似尺度的多参考视点选择方法。实验结果证明，本文方法能在保证编码质量的前提下，减少视差估计的计算复杂度。     

本征图像分解的可变尺度局部分析与集成方法

针对自然图像与磁共振图像,提出本征图像分解的统一的数学模型与算法,解决这两类图像中的重要问题:1)自然图像的光照和反射图像的估计,2)磁共振图像中的偏移场估计与分割.文中数学模型只需要一个基本的假设,即观察到的图像可近似为两个具有不同特性的本征图像的乘积:一个光滑的图像,简称为S-图像;一个近似为分片常量的图像,简称为L-图像.为了充分利用本征图像的特性,提出可变尺度局部分析与集成的方法.由于S-图像的光滑性,使用低阶泰勒展开式或更一般的光滑基函数的线性组合以局部逼近.得到的局部光滑逼近可通过整个感兴趣区域(ROI)的局部区域覆盖及其对应的单位分解扩展成整个ROI上的光滑图像,同时得到图像分割结果和L-图像.实验表明,文中方法对图像的两个本征因子的假设较弱,适用于更广泛的图像.目前方法已在磁共振图像及自然图像中进行测试,得到较优结果.     

基于圆柱面映射的快速图像拼接算法

针对现有基于圆柱面映射的全景图像拼接算法无法实现自动估计焦距的问题,为满足实时性要求,提出了一种基于预测的快速特征点匹配算法,在基于纯旋转运动的自动焦距估计算法基础上,提出了一种基于单应矩阵的焦距修正算法。该算法首先从待拼接图像中提取Harris角点,然后提取方向梯度直方图(Histogram of Oriented Gradient, HOG)描述子进行特征点匹配,基于前一帧图像获得的图像映射关系采用一种基于预测的快速特征点匹配算法,然后使用简化的基于纯旋转运动的焦距估计算法估计出焦距初值,并采用基于单应矩阵的焦距修正算法得到更精确的焦距值,最后将平面图像投影至圆柱平面,使用加权平均融合算法进行拼接,合成全景图像。采用多个测试序列图像对算法进行测试,特征点匹配速度较传统方法提高了10倍以上,自动焦距估计算法能够准确估计摄像机焦距,且耗时仅50毫秒左右。实验结果表明,提出的算法能够快速地合成高质量的全景图像,拼接后的图像畸变小,具有较高的实用价值。     

Higher-order SVD analysis for crowd density estimation

This paper proposes a new method to estimate the crowd density based on the combination of higher-order singular value decomposition (HOSVD) and support vector machine (SVM). We first construct a higher-order tensor with all the images in the training set, and apply HOSVD to obtain a small set of orthonormal basis tensors that can span the principal subspace for all the training images. The coordinate, which best describes an image under this set of orthonormal basis tensors, is computed as the density character vector. Furthermore, a multi-class SVM classifier is designed to classify the extracted density character vectors into different density levels. Compared with traditional methods, we can make significant improvements to crowd density estimation. The experimental results show that the accuracy of our method achieves 96.33%, in which the misclassified images are all concentrated in their neighboring categories.