摘要
水平集算法为了克服噪声对其演化过程的影响,通常在速度停止项中使用高斯滤波器。但高斯滤波器在滤除噪声的同时也模糊了图像的边缘,而水平集是一种以边缘力为演化动力的算法,因此以模糊图像的边缘力为演化动力通常使传统水平集算法产生错误的分割结果。为此,对水平集函数的速度停止项进行了改进。首先,改进后的速度停止项函数使用一种新的高斯滤波器,该滤波器通过在x和y方向上选用不同的高斯尺度,从而使滤波器在去噪时较好地保留图像边缘等重要信息。接着,对新高斯滤波器长轴和短轴的尺度估计方法进行了研究。最后,再将本文的速度停止项带入到水平集演化方程中,对图像进行演化运算。实验结果表明,采用此方法能够使水平集方法获取更好的分割结果。
The speed stop term was used to realize image segmentation in traditional level set method. A standard Gaussian filter is often used in speed stop term to make the level set insensitive to the noise. However, the Gaussian smoothing filters, while smoothing out the noise, also remove genuine high-frequency edge features and degrade the detection of low-contrast edges. The level set is a method that evolved according to the edge force, and the smoothing filters in speed stop term should be improved. In this paper a new Gaussian smoothing filter was researched. Firstly, the new Gaussian smoothing filter chooses different scale in the x and y directions to protect the edge. Secondly, discuss the determination of short axis＇ scale and long axis＇ scale in Gaussian filter. Last using the new Gaussian smoothing filter as the filter of the speed stop term and evolves the level set function. Finally, the experimental results showed that the developed algorithm is verified well than the traditional one.
出处
《宇航学报》
EI
CAS
CSCD
北大核心
2008年第5期1652-1655,1661,共5页
Journal of Astronautics
基金
航空基础科学基金(20070153005)
航空支撑科技基金(07C53007)
