基于自适应阈值的三维图像多尺度边缘检测

目的研究三维图像边缘检测中边缘的定位精度问题,根据多尺度小波变换的方法,提出一种自适应阈值三维图像多尺度边缘检测技术。方法首先对三维医学图像分解成多幅二维图像,再对图像直方图滤波平滑处理,消除噪声的影响,然后对平滑后的直方图进行多尺度分析,找出直方图的谷点,以不同尺度下的谷点比较后自动确定精确阈值,按照阈值对图像分割,然后对分割图像进行边缘检测,最后将多幅二维图像合成三维图像。结果实验表明,该方法能够自动准确选择分割阈值,准确检测三维图像的边缘。结论算法能解决人工估算阈值不够准确的问题,所检测到的三维图像的边缘能够满足目标识别和三维重建的要求。

Multi-scale edge detection of 3D image based on adaptive threshold

Aim To research the accurate location of 3D image edge in edge detection.According to the method of wavelet transform at multi-scale,a technique of 3D medical image edge detection at multi-scale with adaptive threshold is proposed.Methods 3D image is decomposed into 2D images,then histogram of image is smoothed by filter for noise reduction.Then,smoothed histogram is analyzed in multi-scale for finding valley point.Comparing vale points of different scale to assign automatically accurate threshold for segmenting image.Finally,image edges is detected and 2D images are composed to 3D image.Results Experimental results have shown that accurate segmenting threshold can be chosen automatically,and precise 3D image edge can be detected.Conclusion The inaccurate problem of estimating threshold can be solved using the proposed algorithm.Edge detecting result can be satisfied with demands of object recognition and 3D reconstruction.