基于模糊度测度和改进遗传算法的图像分割  

Image Segmentation Based On Fuzziness Measure and Improved Genetic Algorithm

作　　者：郑松峰[1] 孙飞[1] 徐维朴[1] 

机构地区：[1]西安交通大学人工智能与机器人研究所,西安710049

出　　处：《信号处理》2003年第1期15-18,共4页Journal of Signal Processing

摘　　要：为了利用模糊集理论进行图像分割,本文提出了一种模糊度测度函数,它衡量了背景模糊集和某一个二值图像的相似性:模糊度越小,这种相似程度越大;并且对模糊度的计算只需要进行简单的比较运算和加法运算,便于硬件实现。本文利用遗传算法求取最佳阈值,针对遗传算法的某些不足之处.我们利用精英策略和模拟退火的思想,对其做了一些修正。实验结果表明:相对于Shannon熵方法和模糊熵方法,本文方法取得了较好的分割结果,并且具有较强的抗噪声能力。To segment images with fuzzy set theory, we proposed a fuzziness measure, which measures the similarity of a gray image and a binary image, the less this measure, the more similar the gray image and the binary image. It only needs compare and addition operations to compute this fuzziness measure, and this character makes it convenient to be implemented by hardware. We used genetic algorithm (GA) to obtain the best threshold. To overcome some limitations of GA, we amended the standard GA with elite strategy and the thought of simulated annealing. The experimental results indicate that our method can obtain better results compared with the Shannon entropy method and fuzzy entropy method, and our method is more robust to noise.

关 键 词：图像分割 计算机视觉 模糊理论 数字图像处理 模糊度测度函数 遗传算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN911.73[自动化与计算机技术—计算机科学与技术]