虚拟距离窄带活动轮廓模型

针对活动轮廓模型计算量大、演化收敛缓慢、效率低下的问题,提出了一种新颖的活动轮廓模型.该模型采用虚拟的符号距离函数代替真实的符号距离函数,依靠待检测目标物内外均值来驱动活动轮廓的演化,利用虚拟距离函数的梯度形成一个窄带,活动轮廓在窄带内做简单的加减运算演化.其演化具有计算简单、分割效率高、能自由改变拓扑、全局性的优点,初始化曲线也无须非常接近待检测物体的边缘.符号距离函数重新初始化也只需在窄带内使用高斯函数规则化后,对其取符号运算即可.最后给出了活动轮廓在窄带内收敛的一个简单条件,能方便地判断待检测目标是否被检测出来.

Narrow band active contour model based on virtual distance

Disadvantages of the active contour model with high computation cost, slow speed, and especially low efficiency are carefully researched and a novel active contour model is presented. The presented model replaces the real symbol distance function with the virtual symbol distance function, and the evolution of the active contour in presented model depends on the averages inside and outside the object. The gradient of the virtual distance function form a narrow band, where the active contour evolutes by simply adding and simply subtracting. Thus, the evolution has the following advantages: simple calculation, high-efficiency segmentation, free topology, and global property. The virtual symbol distance function is re-initialized with sign function after the virtual symbol distance function is regularized by the Gaussian function within a narrow band. In addition, a simple condition is given for the active contour convergence within the narrow band and it is easy to judge whether the object is detected.