摘要
用Snake模型分割自然背景下的人造目标时,Snake曲线往往被复杂背景所吸引,无法收敛到人造目标的边缘。针对该问题,文章从目标特征的角度,将分形维数特征引入Snake模型。利用自然背景和人造目标在分形维数特征上的显著区别,定义了基于目标分形维数特征的梯度加权函数,来自适应调整图像梯度幅值的大小,抑制自然背景的干扰。同时,该模型允许初始轮廓远离目标的真实边缘,降低了Snake模型对初始位置的依赖性。实验表明,该Snake模型能够克服复杂自然背景的干扰,提取出人造目标的边缘。
This paper proposes a novel snake model based on fractal dimension feature to segment man-made object in nature background. It first uses the fractal dimension feature to define an adaptive weighting function and then introduces it into the snake energy in order to suppress the nature clutter, Snake driven by the proposed model can evolve from the initial position to the man-made object without disturbances of nature clutters, even if the initial curve is far away from the solution. Experimental results validate the effectiveness and feasibility of the model.
出处
《计算机工程》
EI
CAS
CSCD
北大核心
2006年第18期32-34,共3页
Computer Engineering
基金
中国科学院科技创新基金资助项目(A010416)
