摘要
提出了利用多层细胞神经网络实现梯度矢量流GVF场的方法,并与扩展、细化的细胞神经网络(CNN)相结合来实现动态轮廓的图像分割.细胞神经网络具有并行运算的能力,可解决传统串行算法复杂性大,不能实时处理的问题,并克服了梯度场作为CNN的外力驱动方法的局部最小问题.在图像处理过程中,外部图像由GVF信息引导,最后收敛到所期望的目标位置.结果表明,该方法在不同的输入图像条件下均获得了比Vilarino提出的方法更好的分割结果,并具有实时处理速度.
An implementation method of GVF field using multilayer cellular neural networks (CNN) is proposed, which is combined with expanded and thinned model of CNN to realize the image segmentation strategy of active contours. The CNN has parallel processing ability. Therefore it not only solves the problem of computational comlexity of GVF field using traditional serial computation, but also avoids the local minimum in CNN active contours. Guided by GVF information, external image will be evolved until it reaches the desired position. The experiment results show that the proposed method is better than that of Vilarino, et al ,for different input images and it also has real time processing speed.
出处
《复旦学报（自然科学版）》
CAS
CSCD
北大核心
2004年第1期54-61,共8页
Journal of Fudan University：Natural Science
基金
国家自然科学基金资助项目(60171036)
关键词
图像处理
图像分割
细胞神经网络
动态轮廓(蛇)
梯度矢量流
image process
image segmentation
cellular neural networks (CNN)
active contour (Snake)
gradient vector flow(GVF)
