基于高斯小波的多尺度积图像边缘检测算法

图像的边缘是图像最重要的特征之一。由于边缘和噪声都是图像的高频分量,提取的图像边缘总是受到噪声的污染。针对边缘检测中存在的噪声问题,本文根据Mallat快速小波变换算法的思想,提出用高斯函数和其一阶导数分别作为低通和高通滤波器对图像进行多尺度分析。为了精确定位图像边缘,对各尺度的低频、水平、垂直和对角分量不进行下采样。然后提取不同尺度上的系数,利用多尺度积对噪声严重的图像进行边缘检测。最后根据边缘点的梯度方向,采用改进的局部梯度极大值搜索方法获得图像的单像素边缘。实验结果表明本文所提出的方法,能在被噪声污染严重的图像中提取图像的单像素边缘,且边缘图像信噪比高。     

基于Finite Ridgelet变换的图像线性特征提取

线性特征是图像的一种重要局部特征，它常常决定图像中目标的形状。线性特征的提取在图像匹配、目标描述与识别以及运动估计、目标跟踪等领域具有十分重要的意义。常用的线性特征检测方法有Radon变换和Hough变换，但检测曲线复杂度会很高。本文提出一种多尺度几何分析的线性特征检测方法，该方法以finite ridgelet理论为基础，结合正交小波变换对线性特征进行提取。Finite ridgelet变换对于含有直线奇异的多变量函数具有良好的逼近特性，能够获得连续空间函数的稀疏表达，同时具有区域平滑性、很好的可逆性和去冗余性。实验结果表明，本方法即使在背景复杂的环境下也具有良好的检测效果。     

基于形状和纹理的图像检索

构造了Radon变换的不变量，提出一种新的基于形状和纹理的图像检索方法。对小波图像边缘提取Radon不变量作为形状特征，同时提取小波各频道能量作为纹理特征。然后将形状特征和纹理特征分别进行高斯归一化，计算图像形状和纹理的相似度。最后，利用形状和纹理相似度的加权和进行图像检索。试验结果表明该方法对噪声具有较强的鲁棒性，具有尺度、平移和旋转不变性。     

SAR图像道路检测及其仿真

针对图像边缘定位易发生偏移且对噪声比较敏感的问题,传统检测方法有缺点,现从小波理论出发,提出了一种基于beamlet变换的SAR图像道路线特征提取方法。首先对未经斑点噪声抑制的原始SAR图像进行非下采样小波分解,在变换域中利用beamlet变换进行道路线特征提取,最后采用识别算法检测了道路。实验结果显示方法与传统检测算子相比,具有很好的抗噪性能,检测到更多细节特征且道路识别效果更好。     

一种基于Radon变换的图像水印算法

介绍一种基于几何不变域的图像水印算法.在对原始图像进行Radon变换后,再对Radon变换域中的值进行二维DFT变换,得到的量具有抗尺度变换攻击和旋转攻击的特性,即得到一种几何不变域.最后,按照一定的方法在此不变域中嵌入的水印.实验结果表明,该算法不仅能够抵抗尺度变换攻击和旋转攻击,而且对高斯噪声攻击也有鲁棒性.     

基于各向异性高通滤波的图像边缘检测算法

图像在处理的过程中,总会受到噪声的污染.由于噪声和边缘都是图像的高频分量,在滤除噪声的同时,也破坏了图像的边缘.为了使所提取的图像边缘更加逼近被噪声污染的图像真实边缘且定位精确,提出结合各向异性高通滤波和多尺度积对图像进行边缘检测.首先采用具有各向异性的非下采样Contourlet变换(NSCT)对原始图像进行多尺度、多方向分解,并用多尺度积对变换结果的高频分量去噪,最后利用各向异性高通滤波器长轴与边缘方向之间的夹角确定图像边缘.实验结果表明,所提出的方法抗噪声能力强,计算复杂度低,所提取边缘清晰、光滑且定位精确.     

非线性Radon变换及其在人脸识别中的应用

探讨抛物线、双曲线、椭圆3种非线性Radon变换及其性质,分析它们之间的关系,并将所述3种非线性Radon变换应用于人脸识别．通过对这3种非线性Radon变换及其性质研究得出,当抛物线、双曲线及椭圆的形状参数趋于无穷大时,图像抛物线Radon变换与线性Radon变换相等,双曲线Radon变换与椭圆Radon变换相等;同时,非线性Radon变换具有降噪功能和表达图像纹理特征的特点．文中将受噪声污染的人脸图像分别表示为3种非线性Radon变换下的特征矩阵,并结合PCA算法应用于人脸识别．实验结果表明非线性Radon变换在人脸识别中的有效性．     

仿射不变的多尺度自卷积熵提取方法

为从在不同视角获取的同一场景图像中提取更加独特的不变特征,提出一种图像仿射不变特征的提取方法.首先基于多尺度自卷积变换(MSA)构造了一组新的变换量--多尺度自卷积熵(MSAE);并让明了该熵具有仿射不变性;最后利用最小距离分类器分别对视点变换图像,以及加噪声,加部分遮挡视点变换图像进行分类识别实验.实验结果表明,MSAE特征能够获得更高的正确识别率.     

基于Radon和解析Fourier-Mellin变换的尺度与旋转不变目标识别算法

由于正交矩对噪声鲁棒性强、重建效果好,因此被广泛应用于目标识别与分类中,但是正交矩本质上缺乏尺度变换不变性,而且必要的图像二值化与规一化过程会引入重采样与重量化误差。为此,在研究现有正交矩的基础上,提出了一种基于Radon变换和解析Fourier-Mellin变换的尺度与旋转不变的目标识别算法。该算法首先直接对目标灰度图像进行Radon变换,然后对Radon变换结果进行进一步解析,通过Fourier-Mellin变换将原图像的旋转变化转化为相位变化,将原图像的尺度变化转化为幅度变化;最后,通过定义一旋转与尺度不变函数,同时利用不变函数的4种特征,再应用k-近邻法实现分类。理论与实验结果表明,由于避免了正交矩方法存在的重采样与重量化误差,该算法的分类精度高于基于正交矩的分类方法,而且对白噪声的鲁棒性也显著高于基于正交矩的识别与分类方法。     

基于Radon和解析Fourier Mellin变换的尺度

由于正交矩对噪声鲁棒性强、重建效果好,因此被广泛应用于目标识别与分类中,但是正交矩本质上缺乏尺度变换不变性,而且必要的图像二值化与规一化过程会引入重采样与重量化误差。为此,在研究现有正交矩的基础上,提出了一种基于Radon变换和解析FourierMellin变换的尺度与旋转不变的目标识别算法。该算法首先直接对目标灰度图像进行Radon变换,然后对Radon变换结果进行进一步解析,通过FourierMellin变换将原图像的旋转变化转化为相位变化,将原图像的尺度变化转化为幅度变化;最后,通过定义一旋转与尺度不变函数,同时利用不变函数的4种特征,再应用k近邻法实现分类。理论与实验结果表明,由于避免了正交矩方法存在的重采样与重量化误差,该算法的分类精度高于基于正交矩的分类方法,而且对白噪声的鲁棒性也显著高于基于正交矩的识别与分类方法。     

基于样条小波的有噪图像边缘检测方法

图像的边缘和轮廓包含了图像的重要特征信息。在模式识别等领域中,能否正确检测到图像的边缘轮廓是十分重要的。当图像数据混有噪声的时候,对边缘的检测比较困难,文中采用了小波多尺度去噪和局部极大模边缘检测相结合的方法,在受白噪声污染的图像边缘检测方面得到了较好的效果。     

A novel Hough transform method for line detection by enhancing accumulator array

In this paper, an improved Hough transform (HT) method is proposed to robustly detect line segments in images with complicated backgrounds. The work focuses on detecting line segments of distinct lengths, totally independent of prior knowledge of the original image. Based on the characteristics of accumulation distribution obtained by conventional HT, a local operator is implemented to enhance the difference between the accumulation peaks caused by line segments and noise. Through analysis of the effect of the operator, a global threshold is obtained in the histogram of the enhanced accumulator to detect peaks. Experimental results are provided to demonstrate the efficiency and robustness of the proposed method.     

Automatic extraction of main roads using aerial images

In this paper we describe a method for automatic extraction of main roads in high resolution aerial images. Because a road may vary in color we aim at the detection of its centerline that is more uniform in color and shape. For this purpose, the centerline is modeled as being a chain of short line segments. The method is based on an iterative process and is composed by two stages: in the first stage seed segments are detected using the Radon transform. The second step is the automatic extraction of the roads centerlines starting from the seeds that were found in the first stage. The Radon transform is again applied to find the chain/series of segments candidates to centerline of main roads. In the final part of the paper we present experiments using this approach and the results compared with a reference image. The accuracy of the automated method is evaluated considering the completeness, correctness, and RMS indexes.     

Radon变换在虹膜去噪中的应用

Radon变换在数学理论上有重要价值,并可用于图像直线特征的提取。本文基于Radon变换的基本原理,针对在虹膜识别中眼睑和睫毛等噪声信息的存在,提出一种运用Radon变换检测直线的方法检测眼睑并将其去除。通过在MATLAB上仿真结果表明,很好的抑制了噪声对虹膜识别的影响,可达到98.86%的识别率,并在保持1.143%较低拒绝率(FRR,false rejec-tion rate)的同时,也使误识率(FAR,false acceptance rate)减小到0%。     

一种新的线边界检测方法

提出了一种超声图像的多尺度非线性抑噪的自适应线边界检测方法，首先对超声斑点图像进行多尺度分解，然后对高频部分应用非线性软阈值方法来抑制斑点噪声，再利用逆小波变换重建图像，最后采用基于“窄条”的线边界检测方法对降噪图像进行处理，以不同方向与大小的“窄条”来近似组织边界，其中“窄条”的大小由基于区域增长的局部统计特性决定．仿真实验与组织超声图像验证了此方法在抑制斑点噪声与增强边界方面具有良好的性能．     

Analysis of tubular structures in three-dimensional confocal images

Knowledge about the relationship between morphology and the function of neurons is an important instrument in understanding the role that neurons play in information processing in the brain. In paricular, the diameter and length of segments in dendritic arborization are considered to be crucial morphological features. Consequently, accurate detection of morphological features such as centre line position and diameter is a prerequisite to establish this relationship. Accurate detection of neuron morphology from confocal microscope images is hampered by the low signal to noise ratio of the images and the properties of the microscope point spread function (PSF). The size and the anisotropy of the PSF causes feature detection to be biased and orientation dependent. We deal with these problems by utilizing Gaussian image derivatives for feature detection. Gaussian kernels provide for image derivative estimates with low noise sensitivity. Features of interest such as centre line positions and diameter in a tubular neuronal segment of a dendritic tree can be detected by calculating and subsequently utilizing Gaussian image derivatives. For diameter measurement the microscope PSF is incorporated into the derivative calculation. Results on real and simulated confocal images reveal that centre line position and diameter can be estimated accurately and are bias free even under realistic imaging conditions.     

Wrap-around effect removal finite ridgelet transform for multiscale image denoising

Denoising in wavelet domain is an attractive research field. Recently, the research focused on finding some algorithms that can detect line singularities of image while smoothing noise. Based on the finite Radon transform a finite ridgelet transform was derived for this purpose. But the wrap-around effect of the finite Radon transform limits its power greatly. In this paper, a way is pointed out to remove the wrap-around effect of the finite Radon transform. By embedding it into a moving window pyramid, a multiscale image denoising algorithm is developed.     

Preprocessing of multi-line structured light image based on Radon transformation and gray-scale transformation

In the process of the structured light detection, it is affected by many factors such as the light source, object to it-self and environmental noise, which result in poor imaging quality. For this reason, the Radon transformation of multi angle is adopted to get the transform domain image, after obtaining the information of the oblique angle of the structured light stripes. After that the singular points unrelated to the target stripes are eliminated in the transform domain image. Then the processed transform domain image is restored and the noise is eliminated. At the same time, the above-mentioned structured light images are pre-processed by several existing means to eliminate noise. Then compared with the traditional methods, the superiority of the Radon transformation in eliminating noise interference is highlighted by using the reliability evaluation scheme to evaluate the quality of the processed images. Finally, aiming at the other shortcomings in the Radon transformed image and further improving the image quality, the restored images are handled by the gray-scale transformation enhancement to enhance the overall gray level of the image. The experimental results show that the imaging quality of the image processed by the above methods is significantly improved.