摘要
针对显微图像中的微目标检测与识别,提出了一类基于三次B样条函数多尺度小波变换,该变换通过零交叉或模极值检测图像的边缘,在分解和重建时采用一种基于滤波器系数特征的离散快速算法。为除去由噪声引起的虚假边缘,建立了边缘点检测的自适应阈值方法。为实现微目标的识别,采用了一类简单有效的快速聚类算法。最后采用含随机噪声的模拟图像和显微图像进行算法比较与验证,实验结果证明了这些算法的有效性。
Aiming at micro-object＇s edge detection and identification in micro image, a multi-scale wavelet transform algorithm based on Cubic B-Spline was proposed. In the transformation, the zero cross or modular extremes were used to detect the edge of image, and a fast speed discrete algorithm for decomposition and reconstruction was used on the basis of wavelet coefficient. An algorithm based on adaptive threshold was built to eliminate the false edge caused by noise. Meanwhile, a simple but effective algorithm of fast speed clustering was implemented to identify the target in processed image. The algorithms of simulative images with random noise were compared and identified with those micro images. The exnerimental results show that these algorithms are effective.
出处
《自动化仪表》
CAS
2007年第8期29-31,共3页
Process Automation Instrumentation
基金
湖南省科技计划项目(编号:2006GK3074)
关键词
边缘检测
目标识别
小波变换
快速聚类
显微图像
Edge detection Target identification Wavelet transform Fast speed clustering Microimage
