摘要
针对传统水平集算法计算效率低的缺点,提出了基于对数极坐标的快速水平集算法。该算法首先将均匀采样的图像映射到对数极坐标系下,再进行水平集演化运算,从而实现减少数据量、提高算法执行效率的目的,并对直接法、窄带法以及本方法的计算速度作了对比,实验结果表明,在同样获得理想分割结果的情况下,本方法分割尺寸为128×128,256×256,512×512的图像所用时间分别为0.751min,1.549min和6.113min,其计算速度要远快于另外两种方法。
In view of the lower computational eftuiency of level set, a new algorithm was presented to improve the computing speed of level set method, which combined the Log Polar Transform （LPT） with level set method. The property of the LPT is that it can realize the data compression and reduce the data size through non-uniform sampling of the object. So the pending image was projected into the log polar coordinate to implement the level set evolution, in order to reduce computational efforts. Finally, we compared our method with the direct method and the narrow band method. The results show that our method takes the less time to access the data than other two methods while still obtain the ideal segmentation results.
出处
《光电工程》
EI
CAS
CSCD
北大核心
2006年第5期71-75,共5页
Opto-Electronic Engineering
基金
国家航天基金资助项目(N4CH008)
航空科学基金资助项目(04I53067)
武器装备预研基金(51401040204HK03347)
关键词
水平集
对数极坐标
测地线模型
图像分割
偏微分方程
Level set
Log polar transform （LPT）
Geodesic active contour
Image segmentation
Partial difference equation （PDE）
