Automatic defect inspection of patterned thin film transistor-liquid crystal display (TFT-LCD) panels using one-dimensional Fourier reconstruction and wavelet decomposition

Large-sized flat-panel displays have become increasingly important for use in computer monitors and televisions. This paper has considered the problem of automatic visual inspection of micro-defects including pinholes, scratches and particles in patterned thin film transistor-liquid crystal display (TFT-LCD) panel surfaces. For large-sized TFT-LCD panel inspection, high-resolution line scan is demanded. We propose a global one-dimensional (1-D) Fourier-based image reconstruction scheme that directly works on the 1-D line images instead of the traditional two-dimensional area images. The proposed method fully uses the inherent geometric structure of a TFT-LCD panel. It first eliminates the frequency components that represent the periodic pattern of a TFT-LCD line image in the 1-D Fourier spectrum and then back-transforms the 1-D Fourier-domain image to the 1-D spatial domain image using the inverse Fourier transform. The Fourier reconstruction process can effectively remove the patterned background and distinctly preserve local anomalies in the resulting 1-D image. Wavelet decomposition is further applied to remove uneven illumination in the filtered image so that defects can be easily segmented with simple statistical control limits. Experimental results on a number of micro-defects embedded in TFT-LCD panels show that the proposed method can reliably detect various ill-defined defects without designing and measuring the quantitative features of individual defect types.     

Defect inspection of patterned thin film transistor-liquid crystal display panels using a fast sub-image-based singular value decomposition

Thin film transistor-liquid crystal displays (TFT-LCDs) have become increasingly attractive and popular as display devices. A machine vision approach is proposed for automatic inspection of microdefects in patterned TFT-LCD surfaces. The proposed method is based on a global image reconstruction scheme using singular value decomposition. A partition procedure that separates the input image into non-overlapping sub-images is used to reduce the computation time of singular value decomposition. Taking the pixel image as a matrix, the singular values on the decomposed diagonal matrix represent different structural details of the TFT-LCD image. The proposed method first selects the dominant singular values that represent the repetitive orthogonal-line texture of the TFT-LCD surface. It then reconstructs the matrix by excluding the dominant singular values. The resulting image can effectively remove the background texture and preserves anomalies distinctly. The experiments have evaluated a variety of TFT-LCD microdefects including pinholes, scratches, particles and fingerprints at different image resolutions. The experimental results reveal that the proposed method is effective and efficient for microdefect inspection of TFT-LCD panels.     

Method to eliminate the zero spectra in Fourier transform profilometry based on a cost function

To increase the accuracy, speed, and robustness of 3D measurements in Fourier transform profilometry (FTP), this paper introduces a cost function according to the intrinsic features of the amplitude and frequency modulated (AF/M) signal and proposes two new algorithms to eliminate the background components of the fringe pattern based on the proposed cost function. Finally, the standard Fourier transform (FT) is used to calculate the phase of the pattern, which no longer contains background components. The two proposed methods are both data-driven and require no parameter adjustments in advance. Theoretical analysis and 80 experimental results show that the proposed cost function is valid. The results of more than 80 experiments with different types of fringe patterns, different carrier frequencies, and different noise variances with frequency overlap and sudden phase variation show that the proposed two methods are more accurate and robust than the 2D Gabor wavelet transform, the 2D Fan wavelet transform, and the 1D complex Morlet wavelet transform profilometry, and they are approximately 70 times faster than the 1D complex Morlet wavelet transform profilometry.     

An implementation of synthetic aperture focusing technique in frequency domain

A new implementation of a synthetic aperture focusing technique (SAFT) based on concepts used in synthetic aperture radar and sonar is presented in the paper. The algorithm, based on the convolution model of the imaging system developed in frequency domain, accounts for the beam pattern of the finite-sized transducer used in the synthetic aperture. The 2D fast Fourier transform (FFT) is used for the calculation of a 2D spectrum of the ultrasonic data. The spectrum is then interpolated to convert the polar coordinate system used for the acquisition of ultrasonic signals to the rectangular coordinates used for the presentation of imaging results. After compensating the transducer lobe amplitude profile using a Wiener filter, the transformed spectrum is subjected to the 2D inverse Fourier transform to get the time-domain image again. The algorithm is computationally attractive due to the use of 2D FFT. The performance of the proposed frequency-domain algorithm and the classical time-domain SAFT are compared in the paper using simulated and real ultrasonic data.     

Phase-only filtering on the three-dimensional Fourier spectrum of color images

We propose what is to our knowledge a novel technique for filtering a two-dimensional multichannel color image based on its interpretation as a three-dimensional (3D) signal and the application of a 3D Fourier transform. The color image is considered to have the two usual spatial coordinates with the color distribution as the third dimension of the signal. This approach permits the 3D generalization of the usual phase-only filter and phase correlation, and color pattern recognition can be performed by means of a 3D correlation. We analyze two cases: whitening only along the color Fourier spectrum and whitening the whole 3D Fourier spectrum. We study the transformation induced in the color content of the image by the whitening of the color Fourier spectrum and how it affects the correlation distribution. In particular it is demonstrated that the whitening of the 3D Fourier spectrum improves the recognition performance of the correlation.     

Direct fast method for time-limited signal reconstruction

We consider reconstruction of signals by a direct method for the solution of the discrete Fourier system. We note that the reconstruction of a time-limited signal can be simply realized by using only either the real part or the imaginary part of the discrete Fourier transform (DFT) matrix. Therefore, based on the study of the special structure of the real and imaginary parts of the discrete Fourier matrix, we propose a fast direct method for the signal reconstruction problem, which utilizes the numerically truncated singular value decomposition. The method enables us to recover the original signal in a stable way from the frequency information, which may be corrupted by noise and/or some missing data. The classical inverse Fourier transform cannot be applied directly in the latter situation. The pivotal point of the reconstruction is the explicit computation of the singular value decomposition of the real part of the DFT for any order. Numerical experiments for 1D and 2D signal reconstruction and image restoration are given.     

状态形态学滤波与EWT的故障特征提取方法研究

针对往复机械振动信号具有复杂非线性、非平稳等特性,使用一种基于小波框架的自适应经验小波变换和以集合角度处理信号的形态学滤波来进行往复机械故障特征提取。首先使用自适应经验小波变换通过构造尺度空间曲线对傅里叶频谱进行划分,构造合适的正交小波滤波器组以提取具有紧支撑傅里叶频谱的AM-FM成分;然后根据往复机械振动信号冲击性的特点,基于信号本身特性构造形态学结构元素,对提取出的模态进行状态自适应形态学滤波;最后使用多尺度模糊熵对模态进行定量分析并对故障进行识别。将该方法应用到实测数据中,实验结果验证了该方法的有效性,该方法可以准确对往复压缩机气阀故障进行识别。     

A new approach to radionuclide imaging using compressed sensing

Abstract

Single-photon emission computerised tomography (SPECT) and positron emission tomography (PET) are essential medical imaging tools, with inherent drawback of slow data acquisition process. We present a novel compressed sensing-based reconstruction of these images from significantly fewer measurements than traditionally required, thus demonstrating potential of reduction in scan time and radiopharmaceutical doze with benefits for patients and health care economics. Our work effectively shows that high fidelity two-dimensional (2D) SPECT/PET image is reconstructed using compressive sensing with considerably reduced numbers of samples in acquisition stage. The reconstruction of tomographic images is realised by compressed sensing the 2D Fourier projections of k-space data. These 2D projections being sparse in transform domain need fewer samples in k-space and are reconstructed without loss of fidelity. These undersampled Fourier projections can then be backprojected by employing the iterative reconstruction approach for a complete three-dimensional (3D) volume. Compressed sensing of a phantom image and PET bone scintigraphy with radial Fourier samples are performed. The reconstructions of these images are compared to conventionally sampled images using image quality measures like mean square error, peak signal-to-noise ratio and structure similarity (SSIM) index, showing high-quality image reconstruction.     

高频地波雷达目标的识别检测方法

由于高频地波雷达的检测背景十分复杂,目标信噪比特征不明显,使得常规恒虚警检测不能很好地得到背景和目标的估计,针对高频雷达中目标和噪声在三维谱图中的不同特征,应用模式分析策略,提出了基于识别的混合检测结构的检测器（PDPA方法）,并给出了一定的实验结果。     

伪随机阵列和 Harris角测量在视觉检测中的应用

在三维视觉场景重构过程中,一个难题就是场景图像上的坐标点的匹配问题。在工业检测中,通常解决该问题的方法是采用结构光主动视觉技术,如点结构光、线结构光扫描法以及伪随机编码结构光法等。但是要想用一幅图象在三维欧氏空间中重构三维场景,有效的方法是采用伪随机编码结构光照明主动视觉技术。该方法利用编码结构光照明被测场景,利用编码的窗口特性,使场景被测表面每一个特征点都具有唯一的代码而唯一地被辨识。而图像的特征点提取可以采用简单而稳定的Harris角测量。     

基于时空坐标图像的运动点目标检测方法

针对复杂背景下运动点目标检测方法存在的局限性,提出从时间空间角度检测运动目标的方法.在时间空间域里,背景的复杂纹理表现为无限长直线,运动目标则表现为有限长线段.针对该特点,运用逐点灰度均值将背景与运动点目标分离,并根据投影关系将目标投影到二维空间,获得直观的目标运动轨迹.试验结果表明,该方法能较好地从复杂背景中分离出速度不大于1像素/帧的点目标,且可同时进行多目标的检测,抗噪声能力强.     

Passive Detection of Motion Transverse to the Optical Viewing Axis

The passive detection of velocities of objects moving transverse to the optical viewing axis is discussed from the standpoint of the spatial Fourier transform of the image. A simple algorithm is developed which yields the velocity, as a function of time, of an object moving against a complicated background. This algorithm relies on monitoring magnitude and phase of two prominent spatial Fourier components. Measurement of the Fourier components is most simply accomplished with the direct electronic Fourier transform (DEFT) sensor. This device employs coupling between the image intensity, surface acoustic waves, and electron current in a CdS film to decompose the image into Fourier components. Experimental results show strong confirmation between theory and experiment. The paper concludes with a brief comparison of the DEFT implementation versus other systems in terms of speed, sensitivity, and cost.     

Optical Spatial Frequency Filtering of Radiographs

The frequency content of a photographic image can be considerably modified by using a suitable optical system. The Fraunhofer diffraction pattern (the Fourier transform) of a transparency is produced by using parallel light from a laser. This spatial frequency spectrum can be differentially filtered to produce a reconstructed image of arbitrary frequency content. The radiographic image is a convolution of the X-ray source with the object under examination. The filtering operation lessens the degrading effect of penumbra by means of a deconvolution with a filter that is representative of the Fourier transform of the X-ray source.

Manufacture of the spatial filter is described and results showing enhancement of detail in radiographs by reduction of penumbra are presented.     

彩色伪随机阵列在主动三维视觉检测中的应用

要用一幅图像在三维欧氏空间中重构三维场景,尤其是动态的三维场景,常用的点结构光、线结构光扫描法不能满足要求。一个有效的方法是采用伪随机编码结构光照明主动视觉技术:用编码结构光照明被测场景,因为伪随机编码阵列中任一子窗口阵列都唯一,所以通过对场景图像中编码窗口的查寻,可以确定图像中编码点在编码阵列中的位置,从而使场景被测表面每一个特征点都具有唯一的代码,能唯一被辨识,并最终得以重构。     

Speckle photography fringe analysis by the Walsh transform

Two-dimensional Walsh spectral analysis is presented as a new method of numerically processing the Young's fringes diffraction pattern from a double-exposure speckle photograph. The Walsh spectrum of the fringes is more complex than the Fourier spectrum but can be interpreted reliably by further cross correlation of the fringes with square waves. Compared with Fourier spectral analysis, the new technique gives results of almost identical accuracy but with substantially reduced computational effort. The ideas presented have relevance to the general problem of detecting and accurately determining the frequency components of a 2-D sinusoid in the presence of noise.     

Image analysis by fast improved radial harmonic-Fourier moments algorithm

Radial harmonic-Fourier moments (RHFMs) are popular for image reconstruction and invariant pattern recognition due to their properties of translation, scaling and rotation invariant. RHFMs possess lower computation complexity as compared to Zernike moments and Bessel-Fourier moments. However, they always suffer from discontinuity, numerical instability near the center of image, and reconstruction error, especially have a rise for higher order of moments. In this paper, an improvement of radial harmonic-Fourier moments (IRHFMs) is proposed for effectively avoiding the above-mentioned problems.In this paper, a 2D fast Fourier transform algorithm also is applied to the image matrix to obtain the IRHFMs. Simulation experimental results demonstrate the proposed IRHFMs perform better than traditional RHFMs and other classic orthogonal moments including the latest image moments, for example, polar harmonic Fourier moments in terms of the image reconstruction capability and rotation invariant recognition accuracy in noise-free and noisy conditions.     

Estimating 2-D vector velocities using multidimensional spectrum analysis

Wilson (1991) presented an ultrasonic wideband estimator for axial blood flow velocity estimation through the use of the 2-D Fourier transform. It was shown how a single velocity component was concentrated along a line in the 2-D Fourier space, where the slope was given by the axial velocity. Later, it was shown that this approach could also be used for finding the lateral velocity component by also including a lateral sampling. A single velocity component would then be concentrated along a plane in the 3-D Fourier space, tilted according to the 2 velocity components. This paper presents 2 new velocity estimators for finding both the axial and lateral velocity components. The estimators essentially search for the plane in the 3- D Fourier space, where the integrated power spectrum is largest. The first uses the 3-D Fourier transform to find the power spectrum, while the second uses a minimum variance approach. Based on this plane, the axial and lateral velocity components are estimated. Several phantom measurements, for flow-to-depth angles of 60, 75, and 90 degrees, were performed. Multiple parallel lines were beamformed simultaneously, and 2 different receive apodization schemes were tried. The 2 estimators were then applied to the data. The axial velocity component was estimated with an average standard deviation below 2.8% of the peak velocity, while the average standard deviation of the lateral velocity estimates was between 2.0% and 16.4%. The 2 estimators were also tested on in vivo data from a transverse scan of the common carotid artery, showing the potential of the vector velocity estimation method under in vivo conditions.     

Effective bias removal for fringe projection profilometry using the dual-tree complex wavelet transform

When reconstructing the three-dimensional (3D) object height profile using the fringe projection profilometry (FPP) technique, the light intensity reflected from the object surface can yield abruptly changing bias in the captured fringe image, which leads to severe reconstruction error. The traditional approach tries to remove the bias by suppressing the zero spectrum of the fringe image. It is based on the assumption that the aliasing between the frequency spectrum of the bias, which is around the zero frequency, and the frequency spectrum of the fringe is negligible. This, however, is not the case in practice. In this paper, we propose a novel (to our knowledge) technique to eliminate the bias in the fringe image using the dual-tree complex wavelet transform (DT-CWT). The new approach successfully identifies the features of bias, fringe, and noise in the DT-CWT domain, which allows the bias to be effectively extracted from a noisy fringe image. Experimental results show that the proposed algorithm is superior to the traditional methods and facilitates accurate reconstruction of objects' 3D models.     

基于双树复小波和奇异差分谱的齿轮故障诊断研究

针对故障齿轮振动信号的非平稳特征和包含强烈噪声,很难提取故障特征频率的情况,提出了基于双树复小波和奇异差分谱的故障诊断方法。首先将非平稳的故障振动信号通过双树复小波分解为几个不同频段的分量;由于噪声的影响,从各个分量的频谱中难以准确地得到故障频率。然后对包含故障特征的分量构建Hankel矩阵并进行奇异值分解,求奇异值差分谱曲线,确定奇异值个数进行SVD重构降噪,由此实现对故障特征信息的提取。最后再求希尔伯特包络谱,便能准确地得到故障频率。实验结果和工程应用表明,该方法可以有效地提取齿轮的故障特征信息,验证了方法的可行性和有效性。     

一种新的频谱分析图像重建算法及模拟

这种频谱分析图像重建算法利用二维傅里叶级数的各正交基函数的投影向量,去逼近图像 函数的投影向量,利用优化准则,得到各正交基函数的权系数,再经过傅里叶反变换,即可重建出原始的图像函数。计算机模拟表明,该算法重建时间不足1秒,选择合适的近似级数,对复杂场的图像重建,也能使最大误差不超过5%,平均误差不超过1%。