基于高斯混合模型的多区域并行增长图像分割算法

为克服经典区域增长算法中生长规则以及特征选取的困难,提出了基于高斯混合模型的多区域并行区域增长图像分割算法。首先交互选择多个不同区域的种子点,并利用交互式选择的属于每个区域的子块得到混合模型的个数;然后利用最大期望估计混合模型参数作为区域增长的初始参数,并在增长过程中不停地调节模型参数。为了避免初始种子点位置选择对算法性能的影响,采用了多区域并行竞争增长策略。仿真实验获得了较好的分割效果,表明所提出的算法是合理可行的。     

基于混合高斯模型的运动目标检测方法研究

为了能够及时检测到图像场景中的运动目标,提出一种基于混合高斯模型的运动目标检测方法。借助把图像的像素值看成是前景高斯分布和背景高斯分布的组合,进行了背景估计和自适应背景更新;通过对背景差分后的图像进行多目标分割,进行了多个运动目标的检测。实验发现:随着图像序列的背景不断变化,基于混合高斯模型算法能够准确估计出变化的背景,通过将场景图像和背景图像进行差分,检测到感兴趣的运动目标。     

面向语音增强的约束序贯高斯混合模型噪声功率谱估计

提出了一种基于极大似然的噪声对数功率谱估计方法,采用高斯混合模型对每一个频带上的功率谱包络构建统计模型,将时序包络划分为语音和非语音类,它们分别对应于高斯混合模型的两个高斯分量,描述语音和非语音的统计分布,其中非语音高斯分量的均值即为噪声功率谱的最优估计.采用序贯学习的方法,在极大似然准则下逐帧更新模型参数,并逐帧给出噪声功率谱的最优估计值。此外,由于序贯更新过程中语音信号长时缺失,容易导致模型失稳,提出了一种在线的最小描述长度准则(MDL)来判断语音信号是否长时缺失,从而保证了模型的稳定性.实验表明,算法性能整体优于经典的MS和IMCRA算法。     

一种基于高斯混合模型的改进EM算法研究

针对传统EM算法存在估计参数不具有最优性,以及在参数估计中需要人工参与等问题,提出一种基于高斯混合模型的改进EM算法。采用无人工参与的无监督思想,获取高斯混合模型对直方图拟合的最优参数组合。实验表明,该算法不仅能够快速地估计模型参量,而且能够给出最优参数,并在图像增强中使细节更明显,对比度更适中。     

基于混合高斯模型的红外图像自适应校正算法

针对传统时域高通滤波校正算法存在的鬼影问题,提出一种新的基于混合高斯模型的红外图像自适应校正算法。新算法利用混合高斯模型对场景进行建模,只有在像元输出值满足一定条件的时候,才将其更新到校正系数中,实现有选择性地更新校正系数。通过一组仿真和真实的红外图像序列评价算法的性能,仿真图像采用峰值信噪比指标进行定量评价,新算法比传统时域高通滤波校正算法的峰值信噪比提高了约9 dB。真实图像采用主观的定性评价,传统算法校正结果中存在着明显的鬼影,而新算法校正结果中不存在鬼影。     

一种改进的基于混合高斯模型的运动目标检测方法

混合高斯模型方法被广泛应用于运动目标检测,但是现有的混合高斯模型方法在应对噪声和光照突变时效果不佳,其运动目标检测的效果会受到严重影响。为了解决上述问题,提出了一种结合边缘混合高斯模型方法以及改进的基于邻域差方法的综合方法。该方法充分利用了边缘图像对于噪声和光照突变不敏感,以及邻域差方法可以去除部分噪声的特点,对图像序列进行综合处理。实验结果证明,该方法可以提升运动目标的检测率,降低误警率,可以更有效地应对噪声和光照突变的干扰,从而具有更准确的目标检测效果。     

基于改进型高斯混合模型的嵌入式实时运动检测系统

针对在嵌入式CPU上难以做到复杂背景下运动目标的实时检测,提出了一种基于改进型高斯混合模型的实时运动检测方案,采用改进型高斯混合模型,对高斯混合模型进行简化和结构调整,同时进行了C语言层面和CPU层级的优化,使其更合适于嵌入式平台,并详细分析了DM6446平台的软硬件设计,介绍了该算法在DM6446平台上的实现过程。实验结果表明：该系统能够有效克服外界环境变化带来的干扰,能够实时检测,可以实现多目标跟踪。     

SAR影像极化特征的混合高斯模型与分类

针对高分辨率极化合成孔径雷达(SAR)影像中极化特征呈现尖峰拖尾等复杂多样的统计特点,采用混合高斯模型(GMM)对极化特征进行建模,提出了一种约束距离的混合多元高斯分布的参数估计算法。该参数估计算法在贪婪期望最大算法框架下设计约束距离函数,自动估计混合分量的个数和模型参数,进而在贝叶斯框架下实现极化SAR影像的地物分类。对Radarsat-2旧金山等地区三组影像数据的分类结果表明:相比于经典的分类算法,所提GMM分类算法的总体精度提高了7%～10%,且对样本数目的依赖性更小,在城区和耕地区域等异质区域可以得到精度更高的分类结果。     

基于EM和GMM相结合的自适应灰度图像分割算法

提出一种阈值自适应、EM方法估计GMM参量的图像分割算法,能够根据图像的内容结合区域和边界两方面的信息自适应地选择阈值,精确地进行图像边界分割.算法首先提取图像的边界,然后根据边界的直方图计算图像的可分割性,由可分割性确定EM方法的阈值进行GMM分割,最后合并图像的近似区域.实验数据表明,相比其它图像分割算法,以及固定阈值的传统EM算法,本算法的分割结果更为准确.     

基于局部高斯混合特征提取的高光谱图像分类

为了进一步提高高光谱图像的分类精度,提出一种基于局部高斯混合特征提取的分类(LGMFEC)方法.LGMFEC方法首先基于高光谱图像的空间结构为每个样本构建局部邻域集合,然后从局部邻域集合中提取高斯混合特征来充分表征空间-光谱信息及相关变化信息,最后将局部高斯混合特征融入包含黎曼核函数的支持向量机(SVM)分类器中,从而...     

一种双正交心音小波的构造方法

为了提高小波分析在心音信号处理中的性能, 在分析小波构造理论的基础上, 构造了一种专门用于心音信号处理的小波基. 首先提出一种构造滤波器长度为偶数的紧支撑双正交小波的一般方法; 然后根据心音信号的特点, 讨论心音小波的构造原则和一种基于心音 小波族的心音信号合成模型, 并且在此基础上构造出心音小 波. 为了突出使用心音小波处理心音信号的先进性和实用性, 对心音小波进行了比较全面的理论和数值仿真分析. 实验结果表明, 相比常用的db, bior系列小波, 运用心音小波对心音信号进行处理, 能够获得更好的去噪效果、 更精确的心音分类信息以及更小的重构误差率, 为心音特征提取和身份识别的深入研究提供了一种新方法, 在表征心音个体特征的细节方面具有积极的意义. 本文根据应用对象设计专用小波的方法也为工程应用中小波基的选择提供了一种新途径. 关键词： 双正交小波 心音小波 构造方法 心音合成模型     

Computing communities in complex networks using the Dirichlet processing Gaussian mixture model with spectral clustering

Community detection becomes a significant tool for the complex network analysis. The study of the community detection algorithms has received an enormous amount of attention. It is still an open question whether a highly accurate and efficient algorithm is found in most data sets. We propose the Dirichlet Processing Gaussian Mixture Model with Spectral Clustering algorithm for detecting the community structures. The combination of traditional spectral algorithm and new non-parametric Bayesian model provides high accuracy and quality. We compare the proposed algorithm with other existing community detecting algorithms using different real-world data sets and computer-generated synthetic data sets. We show that the proposed algorithm results in high modularity, and better accuracy in a wide range of networks. We find that the proposed algorithm works best for the large size of the data sets.     

复Contourlet域有向图与高斯混合模型的声呐图像增强

提出了复Contourlet域(CCT)中有向图与高斯混合模型的声呐图像增强算法。采用复Contourlet分析提取各尺度中声呐图像每一方向的弱特征信息;为建立特征信息间的联系,考虑复Contourlet域相邻尺度间子带系数的状态具有Markov性,子节点系数的状态依赖于父节点系数状态,构建有向概率图模型反映复系数的这种持续性;尺度内,构建高斯混合模型来建立同尺度中特性信息的联系,以两状态高斯混合模型来表征子带系数的非高斯边缘分布;最后,采用期望最大(EM)算法训练模型参数估计增强图像的系数,实现声呐图像增强。实验结果表明,本文算法与小波域隐马尔可夫树(HMT)算法、Contourlet域HMT算法相比,峰值信噪比(PSNR)增大4 dB以上,结构相似(SSIM)指数增加0.3;本文算法不仅能较好地抑制了声呐图像的强噪声,同时保留了图像边缘和轮廓等弱特征信息。     

Gaussian mixture model-based gradient field reconstruction for infrared image detail enhancement and denoising

Infrared images are characterized by low signal-to-noise ratio and low contrast. Therefore, the edge details are easily immerged in the background and noise, making it much difficult to achieve infrared image edge detail enhancement and denoising. This article proposes a novel method of Gaussian mixture model-based gradient field reconstruction, which enhances image edge details while suppressing noise. First, by analyzing the gradient histogram of noisy infrared image, Gaussian mixture model is adopted to simulate the distribution of the gradient histogram, and divides the image information into three parts corresponding to faint details, noise and the edges of clear targets, respectively. Then, the piecewise function is constructed based on the characteristics of the image to increase gradients of faint details and suppress gradients of noise. Finally, anisotropic diffusion constraint is added while visualizing enhanced image from the transformed gradient field to further suppress noise. The experimental results show that the method possesses unique advantage of effectively enhancing infrared image edge details and suppressing noise as well, compared with the existing methods. In addition, it can be used to effectively enhance other types of images such as the visible and medical images.     

Deep learning-based channel estimation using Gaussian mixture distribution and expectation maximum algorithm

In massive multiple-input multiple-output (MIMO), it is much challenging to obtain accurate channel state information (CSI) after radio frequency (RF) chain reduction due to the high dimensions. With the fast development of machine learning(ML), it is widely acknowledged that ML is an effective method to deal with channel models which are typically unknown and hard to approximate. In this paper, we use the low complexity vector approximate messaging passing (VAMP) algorithm for channel estimation, combined with a deep learning framework for soft threshold shrinkage function training. Furthermore, in order to improve the estimation accuracy of the algorithm for massive MIMO channels, an optimized threshold function is proposed. This function is based on Gaussian mixture (GM) distribution modeling, and the expectation maximum Algorithm (EM Algorithm) is used to recover the channel information in beamspace. This contraction function and deep neural network are improved on the vector approximate messaging algorithm to form a high-precision channel estimation algorithm. Simulation results validate the effectiveness of the proposed network.     

Estimating the number of components in Gaussian mixture models adaptively for medical image

An important but difficult problem of Gaussian mixture models (GMM) for medical image analysis is estimating and testing the number of components by model selection criterion. There are many available methods to estimate the k based on likelihood function. However, some of them need the maximum number of components is known as priori and data is usually over-fitted by them when log-likelihood function is far larger than penalty function. We investigate the log-characteristic function of the GMM to estimate the number of models adaptively for medical image. Our method defines the sum of weighted real parts of all log-characteristic functions of the GMM as a new convergent function and model selection criterion. Our new model criterion makes use of the stability of the sum of weighted real parts of all log-characteristic functions of the GMM when the number of components is larger than the true number of components. The univariate acidity, simulated 2D datasets and real 2D medical images are used to test and experiment results suggest that our method without any priori is more suited for large sample applications than other typical methods.