基于小波变换和余弦变换的Fisher脸识别

结合几种现有的人脸识别特征提取算法,先对人脸图像进行小波分解去噪;然后通过离散余弦变换对低频分量作进一步特征提取和压缩,保留人脸图像中对光照、姿态、表情变化不敏感的识别信息;接着利用PCA和LDA相结合得到最终的识别特征,最后采用欧式距离和最近邻分类器识别人脸。实验采用ORL标准人脸库验证了这种组合的有效性。     

基于小波变换和离散余弦变换的fisher脸识别

本文结合几种现有的人脸识别特征提取算法,先对人脸图像进行小波分解去噪;然后通过离散余弦变换对低频分量作进一步特征提取和压缩,保留人脸图像中对光照、姿态、表情变化不敏感的识别信息;接着利用PCA和LDA相结合得到最终的识别特征;最后采用欧式距离和最近邻分类器识别人脸。实验采用ORL标准人脸库验证了这种组合的有效性。     

基于离散小波变换和ICA支持向量机的人脸识别

为了实现对具有不同光照、姿势和噪声的人脸进行识别并提高识别精度,设计了一种基于离散小波变换和最小二乘支持向量机的人脸识别方法。首先,采用二维离散小波变换对人脸图像进行压缩和降噪,以提取低频特征信息分量,然后采用快速独立成分分析法ICA对经过离散小波变换后的人脸低频分量进行特征提取,以进一步减少人脸特征向量维数。在获取图像特征向量的基础上,采用径向基函数作为核函数,将训练样本数据输入最小二乘支持向量机进行训练以获得最终的分类模型。在ORL数据库下采用MATLAB仿真工具进行仿真,实验结果表明,该方法能有效地实现对人脸识别,与其他方法相比具有较高的识别精度。     

基于Curvelet变换的人脸识别技术研究

提出了一种基于二代Curvelet变换的人脸识别算法,先对人脸图像预处理及Curvelet变换系数的特征提取;然后结合PCA L、DA进行降维处理;再对降维后的系数进行分类,进而对人脸进行识别。为研究表情及光照对识别率的影响,分别采用了ORL与Yale人脸数据库,将该算法与小波变换的算法进行对比。实验结果表明,该算法得到更高的识别率。     

基于BP神经网络的人脸识别系统研究

本文设计了一种基于BP神经网络的人脸识别系统,并对其进行了性能分析.该系统首先利用离散小波变换获取包含人脸图像大部分原始信息的低频分量,对图像数据进行降维;再由PCA算法对人脸图像进行主成分特征提取,进一步降低图像数据的处理量;最后使用经过训练后的BP神经网络对待测人脸进行分类识别.详细介绍了离散小波变换、PCA特征提取以及BP神经网络分类设计.通过系统仿真实验与分析发现:人脸特征的提取是该系统的关键;同时,由于人脸灰度信息的统计特征与有监督训练BP神经网络分类器,使该系统只在固定类别,并且光照均匀的人脸识别应用场景中具有较高的识别准确率.因此,很难在复杂环境中应用.     

基于离散余弦变换的人脸识别方法

人脸图像的庞大信息量使其不适合于直接识别。提出用离散余弦变换处理人脸原始图像,得到降维的特征矢量,并结合前馈神经网络对人脸进行分类识别。通过对ORL人脸库多幅人脸图像的仿真实验表明,系统的识别率较高,且训练时间大大降低,是一种高效的识别方法。     

一种2DDCT与压缩感知结合的人脸识别

针对压缩感知(Compressed Sensing,CS)方法需将图像矩阵转化为向量后进行特征提取,导致数据维数很大,计算复杂等缺点,提出二维离散余弦变换(2DDCT)和压缩感知(Compressed Sensing,CS)相结合的人脸识别方法。新方法首先利用2DDCT将图像变换到频域,压缩人脸图像以去掉人眼不敏感的中频分量与高频分量,这样有效降低了所需特征的维数,减少了计算量;然后通过感知算法进行特征提取得到人脸识别特征,最后运用最近邻分类器完成人脸的识别。在ORL、Yale及Feret人脸数据库的实验结果证明了该算法的有效性与稳健性,特别是在YaleB人脸数据库运用该方法得到了很好的试验结果。     

基于Gabor小波变换和两次DCT的人脸表情识别

提出了一种基于Gabor小波变换和两次离散余弦变换(DCT)相结合的人脸表情特征提取方法,在保留有效的纹理信息基础上降低了表情特征的维数.首先对人脸表情图像进行第一次DCT压缩图像,然后对处理过的图像执行Gabor变换,提取表情特征,进而对得到的不同尺度和方向的特征图像进行第二次DCT,得到包含大量表情信息的低维特征向量,最后用BP神经网络对特征向量分类.实验结果显示该方法识别率较高.     

基于迹变换和旋转增量调制特征的模糊人脸识别

为提高各种不同特殊场景下的模糊人脸识别精确性和鲁棒性,本文提出一种基于迹变换和旋转增量调制编码的特征提取方法.该方法将通信语音编码技术与图像变换技术相结合,首先通过迹线旋转扫描整幅图像,并对迹线上的采样信息进行增量调制编码,从而获得多角度的全局有序结构特征,然后用支持向量机对样本图像的这些特征进行训练以分类并识别图像.实验结果表明,在各种不同模糊级的低质量人脸数据库上,本文方法对不同光照变化、不同拍摄角度、不同遮挡等不同场景的人脸图像均能取得较好的识别效果,与一些传统方法相比识别性能大幅提升,相对于VGGNet和Sphereface两种先进方法在三组不同模糊度测试图像集的平均识别率分别提高2.18％和2.20％,具有更高的识别精度和更好的鲁棒性.     

基于概率神经网络优化DWT-2DLDA的高噪声人脸识别

针对人脸识别中的原始图像存在噪声而影响识别性能的问题,提出了基于概率神经网络优化二维子空间分析的人脸识别方法。首先,使用离散小波变化对图像进行预处理;然后,利用二维线性判别分析进行特征提取;最后,利用概率神经网络完成人脸分类。在ORL和FEI量大通用人脸数据库及自己搜集的数据库上的实验结果表明,在添加噪声的情况下,识别率也可高达98.9%,相比几种较新的识别方法,本文方法取得更好的识别性能。     

Suboptimality of the Karhunen-Loeve transform for transform coding

We examine the performance of the Karhunen-Loeve transform (KLT) for transform coding applications. The KLT has long been viewed as the best available block transform for a system that orthogonally transforms a vector source, scalar quantizes the components of the transformed vector using optimal bit allocation, and then inverse transforms the vector. This paper treats fixed-rate and variable-rate transform codes of non-Gaussian sources. The fixed-rate approach uses an optimal fixed-rate scalar quantizer to describe the transform coefficients; the variable-rate approach uses a uniform scalar quantizer followed by an optimal entropy code, and each quantized component is encoded separately. Earlier work shows that for the variable-rate case there exist sources on which the KLT is not unique and the optimal quantization and coding stage matched to a "worst" KLT yields performance as much as 1.5 dB worse than the optimal quantization and coding stage matched to a "best" KLT. In this paper, we strengthen that result to show that in both the fixed-rate and the variable-rate coding frameworks there exist sources for which the performance penalty for using a "worst" KLT can be made arbitrarily large. Further, we demonstrate in both frameworks that there exist sources for which even a best KLT gives suboptimal performance. Finally, we show that even for vector sources where the KLT yields independent coefficients, the KLT can be suboptimal for fixed-rate coding.     

Multidimensional polynomial transform algorithm formultidimensional discrete W transform

The multidimensional (MD) polynomial transform is used to convert the MD W transform (MDDWT) into a series of one-dimensional (1-D) W transforms (DWTs). Thus, a new polynomial transform algorithm for the MDDWT is obtained. The algorithm needs no operations on complex data. The number of multiplications for computing an r-dimensional DWT is only 1 times that of the commonly used row-column method. The number of additions is also reduced considerably     

Harmonic transform

A new method of spectrum analysis, harmonic transform, is proposed to provide an impulse-train spectrum for signals that are comprised of time-varying harmonics. The essential concept of the transform and some of its properties are discussed. Based on the new transform, a short time harmonic transform is developed for time-frequency analysis. Simulation examples dealing with voiced speech show that the harmonic transform can provide a better frequency representation for signals that contain time-varying harmonics.     

小波变换在傅立叶变换轮廓术中的应用

从小波变换本质、原理出发 ,说明它在提取条纹位相方面的应用 ,通过阐述小波变换与傅立叶变换两者之间的内在联系 ,将小波变换应用到傅立叶变换轮廓术中 ,并着重用傅立叶变换的原理对小波提取位相的原理进行详细解释。模拟结果表明 :小波变换在傅立叶变换轮廓术中的应用是正确的、可行的。     

傅里叶变换与小波变换在信号去噪中的应用

对于高频信号和高频噪声干扰相混叠的信号,采用小波变换去除噪声可以避免用傅里叶变换去噪带来的信号折损。对于噪声频率固定的平稳信号,在对信号进行傅里叶变换后使用滤波器滤除噪声。对高频含噪信号则采用正交小波函数sym4对信号分解到第4层,利用极大极小值原则选择合适的阈值进行软阈值处理,最后利用处理后的小波系数进行重构。实验结果表明,对于高频含噪信号傅里叶去噪会出现严重的信号丢失现象,使用极大极小值原则选择阈值进行小波去噪可以有效地保留高频部分的有用信号。     

Time-varying Fourier transform

A time-varying Fourier transform is defined along with its related power and phase spectra. A convenient recursive technique to compute this transform is also presented.     

Tomography time-frequency transform

The paper shows that the fractional Fourier transform (FRFT) of a signal is the Radon transform of the time-frequency distribution of the same signal. Therefore, a time-frequency distribution known as the tomography time-frequency transform (TTFT) is defined as the inverse Radon transform of the FRFT of the signal. Because the computation of the TTFT does not explicitly require any window or kernel function, high resolutions in both the frequency and time domains can be achieved. When the signal contains multiple components, the cross terms can be effectively removed by an adaptive filtering process that is applied on the FRFT rather than the final result. Therefore, distortions made by the filtering process on the desired signal components can be minimized     

Generalized S transform

The generalized S transform (GST), a family of reversible integer-to-integer transforms inspired by the S transform, is proposed. This family of transforms is then studied in detail by considering topics such as GST parameter calculation, the effects of using different rounding operators in the GST, and the relationship between the GST and the lifting scheme. Some examples of specific transforms in the GST family are also given. In particular, a new transform in this family is introduced, and its practical utility is demonstrated.     

Fractional Laplace transform

A new definition of the fractional Laplace transform is proposed as a special case of the complex linear canonical transform. The proposed fractional Laplace transform reduces to the conventional bilateral Laplace transform and the fractional Fourier transform exactly and hence is better suited for the definition of the fractional Laplace transform as compared to the other definitions proposed earlier in the literature. The advantage of the proposed transform as compared to the conventional Laplace transform lies in providing a free parameter which can be effectively exploited in the filtering and signal separation problems.     

Fast computation of discrete cosine transform through fast Hartley transform

A relationship between the discrete cosine transform (DCT) and the discrete Hartley transform (DHT) is derived. It leads to a new fast and numerically stable algorithm for the DCT.