图象分割阈值平面的进化优化方法

Qptimization of Segmentation Threshold Surface Using Evolutionary Computation

提出了一种基于进化计算的图象分割方法。它首先检测图象的边界,将已经检测得到的边界作为图象分割阈值平面的插值限制条件,将图象分割转化对这种分割阈值平面的插值优化问题。采用浮点数编码的进化规化方法进行计算,进化规划的适应度函数由插值平滑性、分割区域的结构有效性和图象边界点阈值勤不变性个限制条件决定,在进货规划收敛时得到图象分割阈值平面。对不同图象的分割实验显示了3这种方法的有效性、合理性和较强的抗噪声干扰能力。

An image segmentation method using evolutionary computation technology is proposed in this paper. It turns the image segmentation into an optimization problem of interpolating threshold surface. The image edge map is first detected, then an optimal threshold surface for segenting image into different areas is optimized based on evolutionary strategies with float encoded individuals. The fitness function of the evolutionary strategies is determined by the constraints of smoothness, validation of segmentation and the invariance of threshold at edge opints. The experimental result has showsn its good performacnce of segmentation efficiency, validation and insensitivity to noise.