摘要
针对传统分水岭算法存在的过分割和对噪声敏感问题,提出了一种能很好地抑制噪声、剔除图像的伪边缘、准确定位图像边缘信息的方法。首先采用高频强调滤波对梯度图像进行增强,然后利用B样条函数对增强后的图像进行多次拟合,最后对拟合的曲面进行分水岭分割。实验结果表明,通过该法处理的梯度图像再进行分水岭变换,有效避免了过度分割问题;同时准确定位了图像边缘信息,提高了分割精度。
With regard to the over-segmentation of traditional watershed algorithm and the problems of sensitivity to noise, a new algorithm that could effectively restrain noise, eliminate image edges and detect the image edges exactly was presented. Firstly, morphological gradient image edge was enhanced by high frequency emphasize filter effectively. Then, the enhancement image was fit by B-spline function many times. Finally, watershed segmentation was used for the smoothing surface. Experimental results demonstrated that watershed segmentation with the proposed algorithm resolved the phenomena of over-segmentation well. The scheme could detect the image edges exact- ly and improve segmentation precision.
出处
《压电与声光》
CSCD
北大核心
2008年第3期356-358,共3页
Piezoelectrics & Acoustooptics
基金
重庆市自然科学基金计划重点资助项目(CSTC2007BA2023)
重庆市教育委员会科学技术研究基金资助项目(KJ070620)
关键词
高频强调滤波
B样条函数
曲面拟合
分水岭算法
图像分割
high frequency emphasize filter
B-spline function
surface fitting
watershed algorithm
image segmentation
