悬浮式生物芯片检测中的图像处理研究

介绍了悬浮式生物芯片的特点和检测方法,首次提出了完整的悬浮式生物芯片图像处理解决方案。使用预处理去除图像中的噪声并对图像予以增强,然后通过模式匹配技术识别样点,确定位置坐标,定位误差小于1个像素。用数学形态学方法圈定样点范围,并用掩模技术提取信号强度值。使用整体或局部法处理背景。结合图形化开发环境LabVIEW和IMAQ工具包进行整个图像处理的集成和实现。结果表明,该方法能有效完成悬浮式生物芯片检测中的图像处理任务。     

生物芯片图像微阵列偏转角度计算及样点分割算法

本文首先介绍了样点分割在生物芯片技术中所处的重要意义,并分析了生物芯片样点偏转角度误差来源以及构成原因,提出了全自动化的图像微阵列偏转角度计算及样点分割的统一解决方案.算法依据生物芯片样点阵列空间分布特点,并基于重构原理,提出了环形投影方法和功率谱估计相结合的功率切片方法,并以此建立起样点阵列的像素位置、偏转角以及功率值的三维分布关系,实现斑点阵列中偏转角度的精确计算.在获得偏转角度值的条件下,采用定向投影法和功率谱估计方法,计算样点阵列的行列间距;利用行、列间距,在一维灰度和序列中实现分段搜索,获取光斑中心粗定位;利用轮廓图像投影后出现双次峰的现象,估算样点直径;最后,依据重心法调整斑点中心,同时采用邻域搜索算法调整斑点直径,最终实现生物芯片扫描图像的精确定位和分割.     

基于提升格形态小波的生物芯片图像增强

提出一种使用提升格形态小波进行生物芯片图像滤波增强的方法。根据生物芯片图像的样点和噪声区域的大小选择合适的结构元素或者预测-升级算子,并通过形态学算子或者提升格构造形态小波分解和重构形式。利用形态小波的不同级连方式和高频系数的处理实现生物芯片图像的滤波增强。实验表明,该方法可以有效地结合形态学和小波滤波的优势,降低了运算量,取得良好的生物芯片图像增强效果。     

采用细胞神经网络快速校正生物芯片图像倾斜

倾斜校正是全自动生物芯片图像处理必不可少的环节。针对以往倾斜校正算法耗时过长,本文提出一种快速倾斜校正算法。算法首先用细胞神经网络（CNN）寻求各个样点的中心,然后进行Hough变换,再计算方差来寻求倾斜角,最后利用CNN灰度图像旋转模板进行倾斜校正。本算法利用了细胞神经网络并行处理的特性,并充分考虑了生物芯片图像的特点。理论分析与试验结果显示本文算法能够准确高速地完成倾斜校正。     

序列图像中运动目标检测

提出动态背景下序列图像中的运动目标检测算法。利用像素邻域的各向同性对图像进行归一化,消除亮度变化等因素的影响;利用光流信息并结合小波变换由粗及精计算速度场来配准图像;用当前帧作参考图像,通过时域积分校正背景图像。当前帧与校正后背景图像作差得到差分图像。假设该差分图像中噪声分布为高斯分布,由高斯分布的3σ特性滤除差分图像中的噪声,则粗定位出目标;最后以聚类方法确定运动目标区域。分别对200帧可见光和200帧红外图像序列进行实验,检测率分别为95%和94%。     

基于图像的生物芯片点样质量检测方法研究

目的为了提高生产线上生物芯片点样质量检测的精度与效率,研究基于图像处理和卷积神经网络的算法,判断某生物芯片点样质量是否合格,并检测点样合格的生物芯片上的点样点半径。方法采用CCD相机获取生物芯片点样后的图像,通过图像预处理,利用canny边缘检测和圆的拟合等图像处理方法,得到点样点的几何信息,进而计算出点样点半径。同时提出基于卷积神经网络的点样质量检测方法,通过区域建议网络提取点样点卷积特征,引入分割全连接层来训练检测模型,通过离线训练来验证获得模型的最佳参数。结果和手动测量结果进行对比发现,半径误差不超过±0.1 mm,点样质量检测准确率为91.1%,单个生物芯片检测时间总和不超过1.6s。结论所提出的方法能够满足生产线上产品检测准确性和实时性的要求。     

基于最优模型残差的指纹方向校正法

指纹的方向信息是指纹图像处理中非常重要的信息,指纹方向场的获取以及校正直接影响指纹自动识别的性能。本文提出了一种基于指纹方向场数学模型的方向校正方法,根据指纹中的奇异点性质和分布,建立起与对应指纹方向场的最优方向场模型;然后,根据模型与实际指纹方向场的残差对指纹方向进行校正。由于算法利用指纹整体的拓扑信息对指纹噪声方向进行校正,具有很强的抗干扰能力,较适合劣质指纹图像处理。     

反射式拼接CCD相机渐晕图像校正

为解决反射式拼接CCD相机成像渐晕问题,依据辐射定标理论采用多亮度点渐晕图像校正算法处理渐晕图像。首先,针对典型的反射镜拼接方式,分析了成像时产生渐晕的原因;然后,根据实验室多亮度等级下成像结果,得出拼接CCD渐晕区各像元数码输出值与入瞳处辐射亮度成线性关系的结论;最后,应用多亮度点校正算法,计算出渐晕区各像元的渐晕校正因子,校正渐晕图像。实验结果表明:对非均匀性大于10%的多幅原始渐晕拼接图像,校正后,图像的非均匀性下降到0.5%以内,可满足反射镜拼接CCD相机非均匀性指标要求。     

一种基于Harris特征点和DWT-SVD的图像盲水印算法

目的针对第2代数字水印技术,提出一种基于Harris特征点和DWT-SVD的图像盲水印算法。方法提取归一化图像的Harris特征点;选取部分稳定特征点来确定要嵌入水印的特征区域;将特征区域作一次小波分解得到的低频子带,对低频子带进行分块,并对每一块进行奇异值分解,通过对每块中最大奇异值进行加权的方法来嵌入水印信息。结果 PSNR值均大于45 d B,NC值接近于1,说明该算法具有可行性。结论该算法对剪切攻击具有很好的鲁棒性,同时该算法也能很好地抵抗噪声、中值滤波攻击、提高亮度攻击、降低亮度攻击、基本图像处理操作的攻击。     

基于局域最大亮度的靶状激光图像边缘检测

针对工程应用中靶状激光图像带有较强噪声干扰和要求图像处理速度快的特点,本文提出基于局域最大亮度的靶状激光图像边缘检测方法.该方法根据靶状激光图像由多个明暗相间的圆环组成的特征,采用一个大小适当的圆形检测模板,让模板沿着亮环按一定规则在一定范围内移动,计算模板内各像素点的亮度和,使亮度和为最大值的模板的中心点为这个区域的边缘点.用最小二乘法将由此方法获得的边缘点拟合定中.用摄像机(640像素×480像素,光靶分辨率为0.0504mm/像素)距光源10m处拍取50幅图像,处理后的定中标准差为0.0237mm.实验和应用结果表明,此方法速度快,精度高且抗噪声性能好.     

Microarray image enhancement by denoising using stationary wavelet transform

Microarray imaging is considered an important tool for large scale analysis of gene expression. The accuracy of the gene expression depends on the experiment itself and further image processing. It's well known that the noises introduced during the experiment will greatly affect the accuracy of the gene expression. How to eliminate the effect of the noise constitutes a challenging problem in microarray analysis. Traditionally, statistical methods are used to estimate the noises while the microarray images are being processed. In this paper, we present a new approach to deal with the noise inherent in the microarray image processing procedure. That is, to denoise the image noises before further image processing using stationary wavelet transform (SWT). The time invariant characteristic of SWT is particularly useful in image denoising. The testing result on sample microarray images has shown an enhanced image quality. The results also show that it has a superior performance than conventional discrete wavelet transform and widely used adaptive Wiener filter in this procedure.     

Signal intensity correction for multichannel MR images using radon transformation

The purpose of this study is to correct signal intensity at low‐field MRI system with multichannel receiver coils using Radon transformation and filtered backprojection (FBP) method. An open‐type 0.32 T MRI system and a body size phantom were used to acquire the MR images. We used various types of coils from 2‐channels to 4‐channels, which minimized the loss of signal. In the intensity correction process, Radon transform was used for the images of each channel and low‐pass filtering was applied to reduce noise. After that FBP was used for the space transform again from the Radon space to the image space. We also made changes to the projection ranges and their intervals, and then confirmed them to evaluate the optimal parameters. All the intensity corrected results were compared with its original sum‐of‐square (SOS) images, and the corrected images showed more uniform and homogeneous intensities than the images without correction. In addition, these results were also shown in the quantitative values through the signal intensity variations according to the cut view along the horizontal lines of the images. The feasibility of our approach and results for signal intensity correction may be useful and helpful for the researchers of low‐field system with multichannel coils.     

Modelling and quality assessment of spots in digital images — With application to DNA microarrays

Based on a recently developed spot segmentation method, we here present a new approach to modelling of individual spots in digital images, e.g. images of DNA microarrays. From the model parameter estimates and residuals we have developed an expedient approach to automatic quality assessment and identification of corrupted spots. The suggested approach to quality control is shown to give a statistically significant decrease of the variance in gene expression log-ratio estimates for three different DNA microarray datasets.     

MLEM deconvolution of protein X-ray diffraction images based on a multiple-PSF model

In this paper we analyze the degradation of protein X-ray diffraction images by diffuse light distortion (DLD). In order to correct the degradation, a new multiple point spread function (PSF) model is introduced and used to restore X-ray diffraction image data (XRD). Raw PSFs are collected from isolated spots in high-resolution areas on the diffraction patterns which represent the orientation of DLDs. An adaptive ridge regression (ARR) technique is used to remove noise from the raw PSF data. A target Gaussian function is used to model the raw PSFs. A maximum likelihood expectation maximization (MLEM) algorithm combined with a multi-PSF model is employed to restore high intensity, asymmetrical protein X-ray diffraction data. Experimental results using a single and multiple PSFs are presented and discussed. We show that using a multiple PSF model in the deconvolution algorithm improved the quality of the XRD and as a result the spot integration error (/spl chi//sup 2/) and corresponding electron density map are improved.     

A novel approach for high-quality microarray processing using third-dye array visualization technology

Historically, microarray image processing has been technically challenging in obtaining quality gene expression data. After hybridization of Cy3- and Cy5-labeled samples, images are collected and processed to obtain gene expression ratio measurements for each of the elements on the array. The hybridization process often brings in contaminating noise, which can make correct identification of the signal difficult. In addition, spot intensity levels are highly variable due to the expression differences of different genes, and weak spots are often difficult to detect. These conditions are further complicated by inherent irregularities in spot position, shape, and size commonly found on high-density microarrays, making image processing an often labor-intensive task that is difficult to reliably automate. We previously reported a novel third-dye array visualization (TDAV) technology that allows prehybridization visualization and quality control of printed arrays. Here, we present a new microarray image processing approach utilizing TDAV. By incorporating the third-dye image, we show that overall quality of the microarray data is significantly improved, and automation of processing is feasible and reliable. Furthermore, we demonstrate use of the third-dye image to better quality control microarray image analysis. Both the principle and implementation of the approach are presented in detail, with experimental results.     

Using random features of dot-matrix holograms for anticounterfeiting

The images on a dot-matrix hologram contain many two-dimensional (2D) dots with different grating orientations and different grating pitches. Because the zeroth-order light nondiffracted by different grating structures has the same progress direction, the nondiffracted light can be diffracted to a 2D spot spectrum by the 2D dot structure. The 2D spot spectrum depends on grating depths and dot sizes. Although ordinary noises are troublesome for 2D spots, noises caused by special dot arrangement defects or special grating moiré fringes are useful in checking holograms. Since the features of grating depths, dot sizes, dot arrangement defects, and grating moiré fringes can be randomly changed on a case by case basis, 2D spot spectra in different cases are different. The aforementioned random features are used to identify dot-matrix holograms.     

Improvement of digital photoelasticity based on camera response function

Studies on photoelasticity have been conducted by many researchers in recent years, and many equations for photoelastic analysis based on digital images were proposed. While these equations were all presented by the light intensity emitted from the analyzer, pixel values of the digital image were actually used in the real calculations. In this paper, a proposal of using relative light intensity obtained by the camera response function to replace the pixel value for photoelastic analysis was investigated. Generation of isochromatic images based on relative light intensity and pixel value were compared to evaluate the effectiveness of the new approach. The results showed that when relative light intensity was used, the quality of an isochromatic image can be greatly improved both visually and quantitatively. We believe that the technique proposed in this paper can also be used to improve the performance for the other types of photoelastic analysis using digital images.     

Removal of noise in MRI images using a block difference-based filtering approach

Magnetic resonance imaging (MRI) images are frequently sensitive to certain types of noises and artifacts. The denoising of MRI images is essential for improving visual quality and reliability of the quantitative analysis of diagnosis and treatment. In this article, a new block difference-based filtering method is proposed to denoise the MRI images. First, the normal MRI image is degraded by a certain percentage of noise. The block difference between the intensity of the normal and noisy MRI is computed, and then it is compared with the intensity of the blocks of the normal MRI image. Based on the comparison, the pixel weights are updated to each block of the denoised MRI image. Observational results are brought out on the BrainWeb and BraTS datasets and evaluated by performance metrics such as peak signal-to-noise ratio, structural similarity index measures, universal quality index, and root mean square error. The proposed method outperforms the existing denoising filtering techniques.