摘要
为了实现对包含在一个Snake区域内的多个目标进行分割的目标,对传统Snake算法进行了改进,抛弃了与轮廓的弹性力和弯曲力有关的内能,由图像的相对黑像素点数构成图像力。变形中追求图像力的最大值,找到了使Snake收缩变形的方法。针对收缩过程中自身缠绕问题提出了判别方法,这种判别方法也即本研究的分裂方法,从而使Snake具有演化分离的功能,即每一个目标都被一个Snake包围分割。试验结果表明,变形后的模型具有如下的特点:能够分割多目标,图像力能量项物理意义直观清晰,且易于实现,对初始轮廓不太敏感,收敛速度快,具有一定的价值。
In order to implement the goal of the multi-objects segmentation in one snake region work, firstly, an improved snake model algorithm is proposed. In the new snake model algorithm, the internal energy related with the flexibility and power of bending is abandoned and the image power is made up of the quality of black pixel, pursued the max. A method of contraction is found. Secondly, against the question of snake twisting, a judgment method is pointed out. This way is also the way of snake splitting, the snake model possessed of the function of separation. The results of experiment indicate that the improved snake model is a multi-object segmentation algorithm, image energy formulations being conceptually simple and rapidly convergent, not being so sensitive to original snake contour, convergence speed being developed, and being have efficiency.
出处
《计算机工程与设计》
CSCD
北大核心
2009年第19期4455-4457,共3页
Computer Engineering and Design
