图像

A algorithm of dim small targets detection in infrared image is proposed. Image is processed by using median filter and unsharp mask. The infrared image intensity surface is fitted by Haralickd discrete orthogonal polynomial basis function,and then the maximum points are detected on the fitted intensity surface by convolving the image with the second order directional derivative operators.

提出一种红外弱小多目标图像检测方法，图像先经过中值滤波、高提升滤波增强处理，再用基于Haralick离散正交多项式曲面拟合图像，拟合曲面的极大值点即目标点，可以利用导出的二阶方向导数算子与图像进行卷积得到。

To obtain the high quality image of discrete phase distribution, the paper puts forward a new algorithm of image reconstruction, which is based on minimum two multiplications and SIRT iterative technique. The experimental results show that the model is reliable, the accuracy of algorithm is higher, the reconstructed image accords with the actual situation, and the quality of reconstructed image is improved.

为获取高质量的离散相分布图像，根据基于约束的最小二乘法和SIRT算法，提出了一种带有最优修正的新的图像重建算法，实验结果证明具有很高的图像重建精度，图像质量得到改善。

With the development of information industry,document image in digital information management is more and more important.

0引言近年来，随着信息采集技术的不断发展，文档图像在信息的数字化管理中越来越重要，文档图像处理逐渐成了人们的研究热点，例如，中英文文档图像混合识别[1]和基于内容的文档图像倾斜校正[2]等。

In order to overcome the aliasing phenomenon commonly existing in real contourlet transform image denoising, some characters of complex contourlet transform whose structure is a cascading of dual tree complex wavelet and directional filter banks are discussed. It is proved that the transform performs well at division and restraining ability under white Gaussian noise condition. Then, an image denoising algorithm was proposed based on the transform.

为了克服实轮廓波图像消噪后广泛存在的混叠现象，研究了基于双树复小波级联方向滤波器架构的复轮廓波变换图像消噪的若干性质，证明了对于高斯白噪声图像，该变换具有更好的分割能力和抑制能力，并在此基础上提出了一种基于该变换的图像消噪算法。

Material and Methods:From March,2008 to January,2009,250 patients(146 Males and 104 Females,ranged from 18 to 82 years old,mean age was 48 years old) enrolled in the First Affiliated Hospital of Kunming Medical College were performed dual energy scan(Group A,200 patients) and Neuro-DSA scan(Group B,50 patients) using DSCT.The patients were randomly divided into Group A and Group B.Regarding to the justification of scan protocol,100 patients in Group A were selected as sub-group to explore the scan protocol.The image quality Was analyzed both in four categories(4.0/60、4.5/60、4.0/70、4.5/70) of flow rate and total amount of contrast-enhanced material and in three categories(100—150HU、150—200HU、200-250HU) of CT value of left common carotid artery.The comparison of two scan approach was based on image quality,radiation does, postprocessing methods,time of scanning and subtraction and data storage amount.To validate the diagnostic capability of DE-CTA virtual unenhanced images,100 patients in Group A were undergoing head scan both in routine non-contrast program and DE-CTA virtual unenhanced program.Mean CT values,signal-to-noise ratio,image quality,sensitivity of lesion identification,and radiation does were applied to comparison.According to diagnostic results of dual energy of cerebral arterial angiography,62 patients out of 200 patients in Group A were confirmed with cerebral vessel lesions,the results were complied with DSA findings(the time between the two methods were within 1 week).

材料与方法收集2008年3月到2009年1月间在昆明医学院第一附属医院行DSCT CTA检查者250例(男性146例，女性104例，年龄范围18-82岁，平均48岁)，随机分为A、B两组，A组200例行双能量头颅CTA扫描；B组50例行Neuro-DSA扫描。A组前100例按对比剂不同流率和总量分为4组(4.0/60、4.5/60、4.0/70、4.5/70)对图像质量比较；按触发时左颈总动脉内平均CT位分3组(100—150HU、150—200HU、200-250HU)，比较图像质量的差异；对A、B两组血管图像质量、辐射剂量、后处理的方法、扫描和后处理时间及数据的容量进行比较分析；A组中前100例行常规与DE-CTA虚拟平扫的平均CT值、信号噪声比、图像质量、病灶显示情况、辐射剂量的比较分析；A组中行DSA检查62例(两种方法间隔时间均1周内)，比较两种检查方法诊断结果是否符合，并对其中24例动脉瘤行动脉瘤部位、数目、形态、瘤颈显示程度、瘤颈大小及瘤体长径和短径的比较研究。

Because of the presence of the difference of the gray value and the distortion between the object image and the reference image, traditional matching method will be degraded. We need more useful characters. Eigenspace approach is a promising candidate for an appearance-based object representation.

文中提出基于奇异值分解的跟踪算法，算法首先建立模板图像训练集合，利用奇异值分解方法，张成模板图像特征空间，然后求出模板图像在特征空间里的投影值，代替传统算法中灰度对两幅待匹配图像进行的全局搜索定位。

Though it can not avoid the problem of image ergodicity, it can decrease the image matching calculation and reduce the computation time of the matching algorithm.

可以通过"子模板"的匹配实现对图像目标的识别与跟踪，虽然同样存在图像匹配遍历问题，但可大大减少图像匹配的计算量，减少图像匹配的计算时间。

The complexity of connectivity search at the worst situation is O, N is the side length of image. The complexity of connectivity ergod is O, N_q is the number of sub-pattern instances.With the connectivity operation of sub-pattern instance, contour extraction, Euler number computing and connected component labeling of binary rectangle NAM image are researched.

以子模式实例的连通关系处理算法为基础，分别研究了基于矩形NAM图像的二值图像轮廓提取、欧拉数计算、连通区域标记等图像运算，给出了具体的算法和分析，并将其与流行的基于其它图像表示的同类相关运算进行了比较。

The instrument software has the function of demonstration of basic knowledge of Digital Image Processing, i.e., it performs the functions of real-time collection and images saving, color change (brightness, binarization, RGB separation etc.), geometrical transform, edge detection (Robert arithmetic operator, Sobel arithmetic operator, Prewitt arithmetic operator, LOG arithmetic operator, Canny arithmetic operator etc.), gray-scale image enhancement (histogram equalization, averaging filtering, median filtering), and binary mathematical morphology management (erosion, dilation, opening, closing, objects marking, Euler number calculation).

该测试仪的软件部分包含数字图像处理课程所涉及的一些基础知识的演示，具体包括图像的实时采集与保存、颜色变换(灰度化、二值化、R、G、B颜色分量的分离等)、几何变换、边缘检测(Robert、Sobel、Prewitt、LOG、Canny)、灰度图像增强(直方图均衡化、均值滤波、中值滤波等)、二值图像操作(腐蚀、膨胀、开启、闭合、对象标注、欧拉数计算)等功能。

Digital image processing technology,the present reseach level about image alarm system is to separate eyewinker from monitoring image,and to judge whether the eyewinker is harmful according to the size of the area of object,this method is imprecise and suspectable.The image recgnition alarm system isan apply on image processing technology.when a foreign object enter monitoring place,we will segment、measure object,and analysis、judge through proportion characer or invariant monent.the result of the parameter will decide whether the eyewinker is harmful,if the object is harmful ,system will auto-startalarm device.

而本设计的主要特点是基于了数字图像处理技术，目前传统图像报警系统的研究水平是将异物从监控图像中分离出来，然后依据异物面积的大小来判断异物是否有害，它对异物有害的判断十分不准确，这种判断结果的有效性是很值得怀疑的；针对于这些弊端，本课题在数字图像处理技术上的应用是对监控场所进行全景监控，如有异物进入，对异物进行分割、测量，从不同的角度进行了分析和判断，进而决定异物是否有害，若满足场景所规定的有害条件，则启动报警装置，发出报警信息。

The split between the two groups can hardly be papered over.

这两个团体间的分歧难以掩饰。

This approach not only encourages a greater number of responses, but minimizes the likelihood of stale groupthink.

这种做法不仅鼓励了更多的反应，而且减少跟风的可能性。