图像识别
2009年6月15日
题目：Human Face Detection Based on SOFM Neural Network
作者：Yang Kun, Zhu Hong, Pan Ying-jie
年份：2006
摘要：Abstract – This paper presented a new method of human face
detection using Kohonen self-organized feature map (SOFM)
neural network (NN). Presently, the human face detection based
on the skin color is one of the main methods, but the most of
color images include a lot of the skin-like information. Therefore,
it is a difficult problem how to effectively extract the skin regions
from the skin-like regions. The paper presented the detection
method based on SOFM Neural Network to solve the problem.
After nonlinear color clustering, skin regions can be roughly
segmented from the skin-like regions, and then human face
regions can be accurately located by the morphology and the
human face features. The experimental results show that the
method can accurately detect human face in static color images
and has stronger robustness to various skin colors.
Index Terms - Human face detection, SOFM NN, Color
clustering, Morphology
该论文暂不提供下载服务 |
题目：一种快速检测光滑半球表面缺陷的方法
作者：乐静 郭俊杰 朱虹 方海燕 邵伟
年份：2004
摘要：针对具有镜面反射特性的光滑球面，提出了一种利用反光带的镜像变形来对表面缺陷进行快速检测的新方法。这种方法通过漫反射反光带增强图像上的缺陷信息，克服了背景噪声在图像处理中会造成较大误差的缺点，也降低了照明系统的设计难度。根据缺陷引起的反光带镜像变形获取缺陷特征，使图像处理软件的工作量大大减少。通过实验，给出了缺陷尺寸和扫描间隔等因素对检测结果影响的分布曲线，为合理提取缺陷尺寸提供了参考。实验表明，该方法的鲁棒性好、效率高，能识别0.5mm以上的全部缺陷，定位误差在+1。之内。
关键词：表面光洁度测量， 球面， 图像测量， 图像分析
文件大小：1.41MB
 |
题目：CCD离散性对面积测量的影响
作者：乐静 郭俊杰 朱虹 刘雅丽
年份：2004
摘要：分析了电荷藕合器件（CCD）在空间和彩色（或灰度）的离散特性对面积测量的影响。通过统计特定区域的边界点的分析图像分割的效果，提出了一种评价CCD相关误差的℃℃方法。举例说明了图像分割方法对面积测量很大，为此，讨论了正确使用CCD和提高面积测量精度的相应措施。
关键词：电荷藕合器件， 面积测量， 误差分析
文件大小：904K
|
题目：Genetic Algorithm for Edge Extraction of Glomerulus Area
作者：Jun Zhang, Hong Zhu and XueMing Qian
年份：2004
摘要：The automation analysis of kidney-tissue image is an important subsystem in the computer aided diagnosis system of kidney disease. In this subsystem, the correct extraction of glomerulus is an important premise to the exact analysis of kidney-tissue image. A glomerulus edge extraction method based in genetic algorithm (GA) is proposed by considering complex characteristics of the image. Firstly, different scale binary images are obtained by adjusting the parameters of LOG filter. Secondly, the crude spline curve fitting for the glomerulus area boundary is got by genetic algorithm based on the small-scale binary-scale binary image. Thirdly, elaborate adjustment of spline fitting curve is performed according to more boundary information of large-scale binary image to get the optimal spline curve. Finally, the glomerulus area can be extracted correctly according to the fitting curve. Experimental result indicates high precision of glomerulus edge extraction from the kidney-tissue image.
关键词：Glomerulus, GA, LOG filter, Region extraction, Spline function
文件大小：1.09MB
|
下一页 |
|
|
