一种快速医学图像亚像素边缘检测方法

针对目前医学图像配准中使用的多是配准精度较低的像素级的边缘检测的问题,提出一种新算法将像素级算子sobel,log,prewitt,roberts,zerocross和canny分别与Zernike矩算子及搜索算法相结合,并对边缘检测结果进行比较,从中选取较优的算子来实现医学图像亚像素边缘检测.实验结果表明该算法克服了像素级算子和Zernike矩算子的缺点,提取的边缘定位精度高,计算时间短,抗干扰性强.     

光斑质心亚像素定位误差的实验研究

根据误差的来源将光斑质心亚像素定位误差归类为随机误差和系统误差,提出一种简单有效的实验方法对光斑质心定位误差进行定量测试.利用高精度一维电动平移台、POINT GREY Flea2-14S3 CCD相机和LED光源构建了测试系统,对测试结果进行研究和讨论,发现了测试系统LED光斑质心定位系统误差的周期性变化规律,计算得到了基于Flea2-14S3 CCD相机的光斑质心定位随机误差为0.018 pixels,系统误差为0.06 pixels,总体误差为0.063 pixels(约1/15 pixels),能够应用于以光斑质心检测为手段的测量系统中.实验表明,该测试系统可以作为估算光斑质心定位误差大小的一种有效的手段.     

Fast subpixel registration method for InSAR images

In conventional registration methods, image interpolation is used to extract subpixel offsets which lead to the limitation that the registration accuracy is restricted by the interpolation unit. Moreover, the computational burden is heavy when high accuracy is demanded. In order to reduce the computational load, we propose an efficient offset estimation method for interferometric synthetic aperture radar (InSAR) image subpixel registration. Firstly, a novel cost function continuously varying with offsets is established by integrating the empirical signal-to-noise ratio (SNR) and the interpolation operation. This suggests a more accurate registration since the accuracy of the estimated offsets does not depend on the interpolation unit. Secondly, an efficient bi-iterative algorithm is employed to solve the cost function in the continuous domain. The subpixel offsets associated with the maximum of the cost function can be exactly obtained with low computational complexity. Simulated and real data are tested to illustrate the good performance and computational efficiency of the proposed method.     

基于Zernike矩和PSO算法的摄像机神经网络标定

提出了一种新的基于Zernike矩和粒子群(PSO)算法的摄像机BP神经网络标定方法。首先,利用Zernike矩和曲率不变性求取圆形标定模板中心的亚像素坐标,提高神经网络训练数据的精度;其次,利用PSO算法优化网络的初始权重和阈值,提高网络的收敛速度和泛化能力。实验结果表明,该方法在X轴和Y轴方向的测量误差小于0.06 mm,整个测试集均方根误差为0.194 mm,证明了该方法的有效性。     

多投影显示系统抗光线干扰算法研究

为了解决多通道投影显示系统中由于光线相互干扰而引起显示墙上图像画质下降的问题,提出了两种抗光线干扰算法。1.基于点对点的抗光线干扰算法:首先建立拍摄相机与投影机之间的亚像素级对应关系,然后依据该对应关系和实验测得的亮度饱和度增益表对输出图像进行亮度衰减和饱和度提升;2.基于图像增强的抗光线干扰算法:综合利用多种图像增强算法来提升图像画质,间接降低光线干扰。针对这两种算法,给出了算法的理论推导并以穹顶24通道投影显示系统为例,进行了实验验证。实验结果表明：两种抗光线干扰算法可有效降低光线干扰和提升图像画质。     

基于Zernike正交矩的图像亚像素边缘检测算法改进

介绍了Zernike矩及基于Zernike矩的图像亚像素边缘检测原理, 针对Ghosal提出的基于Zernike矩的亚像素图像边缘检测算法检测出的图像存在边缘较粗及边缘亚像素定位精度低等不足, 提出了一种改进算法. 推导了7×7 Zernike矩模板系数, 提出一种新的边缘判断依据. 改进的算法能较好检测图像边缘并实现了较高的边缘定位. 最后, 设计了3组不同的实验. 实验结果同Canny算子及Ghosal算法相比, 证明了改进算法的优越性.     

基于改进型SUSAN算子四轮定位参数检测方法实现

针对当前四轮定位仪传感器众多、精度低、操作复杂、维护成本高等缺点,提出一种三维视觉式四轮定位参数检测的新方法,阐述了基于改进型SUSAN算子实现该方法,通过采集随车轮一起运动的棋盘格式目标盘的序列图像,求取车轮旋转轴线的三维方向余弦,进而求得四轮定位几何参数。与传统方法相比具有非接触、实时、操作简便、精度高的优点。     

一种改进的重心法在光带中心提取中的应用

介绍了光切法（LSM）的基本原理,在分析常用的几种光带中心提取法特点的基础上,针对实际光带中光强呈近似高斯分布的情况,提出改进的灰度重心算法。该方法增大了光强大的点在确定光带中心的权重,提高了测量精度,与现有的几种亚像素光带中心提取方法相比,提取的曲线更接近理想值,三维重构的效果很好,通过实验,结果令人满意。     

CCD camera modeling and simulation

In this paper we propose a modeling of an acquisition line made up of a CCD camera, a lens and a frame grabber card. The purpose of this modeling is to simulate the acquisition process in order to obtain images of virtual objects. The response time has to be short enough to permit interactive simulation. All the stages are modelised: in the first phase, we present a geometric model which supplies a point to point transformation that provides, for a space point in the camera field, the corresponding point on the plane of the CCD sensor. The second phase consists of modeling the discrete space which implies passing from the continous known object view to a discrete image, in accordance with the different orgin of the contrast loss. In the third phase, the video signal is reconstituted in order to be sampled by the frame grabber card. The practical results are close to reality when compared to image processing. This tool makes it possible to obtain a short computation time simulation of a vision sensor. This enables interactivity either with the user or with software for the design/simulation of an industrial workshop equipped with a vision system. It makes testing possible and validates the choice of sensor placement and image processing and analysis. Thanks to this simulation tool, we can control perfectly the position of the object image placed under the camera and in this way, we can characterise the performance of subpixel accuracy determining methods for object positioning.