Image Restoration Using Hybrid Features Improvement on Morphological Component Analysis

Images are generally corrupted by impulse noise during acquisition and transmission. Noise deteriorates the quality of images. To remove corruption noise, we propose a hybrid approach to restoring a random noise-corrupted image, including a block matching 3D (BM3D) method, an adaptive non-local mean (ANLM) scheme, and the K-singular value decomposition (K-SVD) algorithm. In the proposed method, we employ the morphological component analysis (MCA) to decompose an image into the texture, structure, and edge parts. Then, the BM3D method, ANLM scheme, and K-SVD algorithm are utilized to eliminate noise in the texture, structure, and edge parts of the image, respectively. Experimental results show that the proposed approach can effectively remove interference random noise in different parts; meanwhile, the deteriorated image is able to be reconstructed well.     

一种新的车牌号码倾斜校正方法

利用矩阵奇异值特征向量计算倾斜车牌图像的旋转角α,把倾斜图像旋转±（90-α）,进行水平倾斜校正;在垂直倾斜校正中,论文提出了K-means聚类拟合直线法和最小二乘拟合直线法等两种校正方法,分别计算出垂直倾斜角θ,并对水平倾斜校正后的图像进行逆错切变换,得到最终的校正图像。实验结果表明,论文方法能快速准确地检测出图像的倾斜角,并具有较强的抗干扰性和应用适应性。相对于Hough变换校正方法,论文算法缩短了处理时间,在相同的条件下,处理时间比Hough变换快了4．13倍。为图像倾斜校正提供了一个新的有效方法。     

基于Contourlet视觉模型和奇异值分解的图像质量评价

针对基于像素值的图像质量评价方法忽视图像结构信息和需要完全参考图像等问题,提出了一种基于Contourlet域奇异值分解CW-SVD,部分参考图像质量评价方法(contourlet weighted singular value decomposition)。在Contourlet域中,利用奇异值向量对图像结构的表征能力,结合人眼视觉敏感性确定每个子带的视觉权重,得到每个子带的评价测度,再综合得出图像的最终评价指标。实验表明,该方法应用于4种类型的降质图像质量评价时,比峰值信噪比(PSNR、MSSIM)等算法具有更好的稳定性和更好的主客观评价一致性。     

Sparse Approximations of the LMMSE Channel Estimation in OFDM with Transmitter Diversity

The linear minimum mean-squared error (LMMSE) channel estimation for orthogonal frequency-division multiplexing (OFDM) systems requires a large number of complex multiplications. We evaluate a simplified LMMSE channel estimation algorithm in a transmit diversity environment by applying a significant weight catching (SWC) technique to the LMMSE fixed weighting matrix. The SWC technique itself is based on modifying the smoothing matrix by leaving the Γ largest values in each row and turning the rest to zeros. This allows the computational complexity of the full LMMSE processor to be reduced by more than 50%. In the well known LMMSE by singular value decomposition (SVD) technique the sparse approximation is accomplished by zeroing out all but the r largest singular values. LMMSE by SVD is the preferred approximation technique for low delay spread channels. However, in channels with large delay spreads, LMMSE by SWC is a better choice in terms of computational complexity and estimation accuracy Igor Tolochkoreceived his Dipl.-Eng. Degree in Electrical Engineering from Polytechnic Institute, Riga, Latvia in 1987 and PhD from Victoria University, Melbourne, Australia in 2005. He was a senior and later principal design engineer in mobile communications at the Riga Semiconductor Institute Alpha (1988 – 1993). During 1993 – 1998, he was involved in research and development activities in communications with different companies in Riga and Melbourne, Australia. From 1998 to 2002, he was with Ericsson Australia as a senior design engineer. Currently, he works for NEC Australia Pty. Ltd. as a senior design engineer in 3G Mobile Department. His current research interests include digital signal processing, indoor and outdoor wireless communications and error control coding. Michael Faulkner(M'84) received the B.Sc. (Eng) from Queen Mary College, London University, UK, in 1970, the M.E. degree from the University of New South Wales, Australia in 1978, and the PhD from University of Technology Sydney in 1993. From 1972 to 1975 he was with STC (now Alcatel) Australia. From 1975 to 1977 he as with the University of New South Wales, and since then as a lecturer and now professor at Victoria University of Technology, Melbourne, Australia where he is director of the Telecommunications and Micro-electronics research centre. Between 1988 and 2000 he spent four periods at Lund University, Sweden. He was co-recipient of the IEE's 1997 IERE prize for a paper on amplifier linearisation. His current interests are, signal processing, radio technology, radio systems and MIMO/OFDM.     

A Novel Sparse Penalty for Singular Value Decomposition

Singular value decomposition (SVD) is a tool widely used in data denoising, matrix approximation, recommendation system, text mining and computer vision. A m a jority of applications prefer sparse singular vectors to capture inherent structures and patterns of the input data so that the results are interpretable. We present a novel penalty for SVD to achieve sparsity. Comparing with the traditional penalties, the proposed penalty is scale, dimen-sional insensitive and bounded between 0 and 1, which are in favor of controlling sparsity. Regulated by the penalty, we provide an efficient algorithm to pro ject a vector onto a given sparse level in O(n) expected time. The efficient pro jection algorithm serve as a drudge for sparse SVD (SSVD). In experiments, SSVD is efficient and could cap-ture the latent structures and patterns of the input data.     

彩色图像DWT变换下的块SVD零水印

为了提高水印技术的鲁棒性,提出了彩色图像离散小波变换(DWT)下的块奇异值分解(SVD)的零水印.首先对原始载体图像进行离散小波变换,然后选择低频子带进行分块,且对每一块进行奇异值分解,水印则由分解得到的最大前m个奇异值产生.实验结果表明,算法对各种攻击有较强的鲁棒性.     

奇异值方法在大口径反射镜面形分析中的应用

为了更好地分析大口径反射镜的面形,引入奇异值方法来分析系统的重力印透造成的大尺度起伏以及由磨削加工、测量噪声等因素造成的中高频误差。首先对奇异值分解的基本方法以及在反射镜表面评价中的具体应用方法进行了研究;之后利用数值仿真,验证了奇异值分解应用在反射镜表面分析中的可行性;最后,将提出的方法应用在实际的反射镜镜面评价之中,得到系统去除高频误差后的结果。所提出的方法对于低信噪比的大口径反射系统面形评价有较好的指导作用。     

基于数据驱动字典和稀疏表示的语音增强

本文提出了一种基于数据驱动字典和过完备稀疏表示的自适应语音增强方法。首先在训练阶段采用干净语音基于K奇异值分解(K singular value decomposition, K SVD)算法训练过完备字典,然后在测试阶段根据含噪语音的噪声方差自适应选择最优的阈值,采用正交匹配追踪算法对含噪语音信号在过完备字典上进行稀疏分解,最后利用系数稀疏表示重构语音信号,从而达到语音增强的目。该方法不像传统语音增强方法那样减少或消去噪声,而是从字典中选取适当的原子表示纯净信号,从而把纯净信号从含噪信号中分离出来。对白噪声和有色噪声环境下重构语音进行了主客观评价。仿真结果显示:该方法能有效去除加性噪声,并且改善了语音质量。      

基于TSVD-SCN的光纤入侵信号识别算法研究

采用随机配置网络(SCN,Stochastic configurat ion network)对光纤振动信号进行识别,常由于光 纤预警系统的背景噪声问题使得网络的隐含层输出接近奇异,直接影响了SCN对光纤数据的 识别准确率。 因此本文提出了一种基于截断奇异值分解(Truncated singular value decomposition,TS VD)的SCN 方法(TSVD-SCN) 对光纤入侵信号进行识别。TSVD-SCN通过对网络的隐含层输出进行SVD分解并设置阈值去 除其中较小的 奇异值,以减少隐含层输出矩阵的条件数,提升网络识别率。本文利用占空比,平均幅差函 数,FFT求能 量占比的方法进行特征提取,采用基于TSVD-SCN算法对不同入侵振动特征矢量进行分类识 别。实验证明, 本文所提算法模型精度比SCN的模型精度更高,可以准确识别光纤入侵信号类型,对SCN网 络在实际应用中对分类精度的提高有着重要意义。     

融合反距离加权和矩阵填充的位置指纹库构建

为解决位置指纹定位在离线阶段构建位置指纹库时耗费的人力和时间成本较大,构建指纹库效率低和利用空间插值法构建的指纹库精度不高的问题,提出了一种融合反距离加权和矩阵填充的位置指纹库构建算法。该算法仅需人工采集定位区域内少量参考点的接收信号强度值用作信标点指纹信息,结合反距离加权算法特性计算出次信标点指纹信息,根据位置指纹库数据矩阵的低秩性,应用奇异值阈值矩阵填充算法构建出位置指纹数据库。仿真实验结果表明,所提算法有效降低了矩阵填充算法构建位置指纹库所需的人工和时间成本,构建出的位置指纹库定位性能优于反距离加权和克里金空间插值法,接近传统人工采集法,显著地提高了位置指纹库的构建效率。