摘要
这里的方法旨在解决复杂多纹理图像中的目标分割问题。沿袭局部化思路,进一步提出基于局部子区域的方法,介绍了该方法下能量函数的设计方法,并与全局统一能量模型方法和局部能量方法进行了实验对比。实验表明,该方法在继承一般局部区域方法的分割能力的同时,较好的解决了复杂多纹理图像的目标分割问题,能较好的分割多纹理形成的物体边界,成功扩展了基于区域的活动轮廓的分割方法对图像的适应能力。
In this paper, a fresh active contours method is proposed to segment complex image with multiple texture blended. By following the localizing thought, a method based on local sub-region is further proposed. So the images could have multiple regions with different statistics model and region sized, rather than assuming that the image conforms to certain global statistics model. Then how to design energy function for the local and sub-localizing method is demonstrated, and the coparison of this method with the global method and the localizing method, additional with initiate place and local radius result. Experiments show that this method is robust and accurate.
出处
《通信技术》
2010年第2期68-70,共3页
Communications Technology
关键词
水平集
活动轮廓
图像分割
多相图
局部化
level set method
active contours
image segmentation
multi-phase image
localizing
