基于多特征活动轮廓的红外图像显著轮廓检测技术
毕业论文关键词：红外图像分割、主动轮廓模型、局部标准差、局部熵、符号控制力函数
Title A infrared image contour detection technique based on multifeatures
Abstract Infrared image segmentation technique has become a research hotspot,and how touse the related characteristics of image to extract the interested part has becomeone of the research direction.In recent years, active contour model is widely usedin image segmentation and has a better effect, there are also some studies thatthe active contour model is applied to the infrared image segmentation, but itssegmentation effect is not very good.In general, the active contour model in theinfrared image segmentation technology is still in exploring stage.In this paper,based on the active contour model and he infrared image with weak edge, seriousheterogeneity, etc, we put forward the new model.According to the characters ofimage pixels,we construct feature vectors to research outline drive direction anddrive combined with the gaussian kernel function. Based on active contour pixels,both within and outside the eigenvector mean and cosine similarity, we optimizethe SPF function instead of the traditional level set evolution equation, anditeration to obtain the final segmentation result.Realize the infrared imagecontour detection significantly.
Keywords: Active contour model, infrared image segmentation, local entropy, localstandard deviation, signed pressure force.
目次
1绪论1
1.1图像分割技术的现状以及传统图像分割方法简介1
1.1.1红外图像的获得及特点1
1.1.2图像分割的地位及定义1
1.1.3传统的图像分割方法3
1.1.4基于主动轮廓的图像分割方法4
1.2论文主要工作及内容4
2相关的数学理论6
2.1曲线演化理论6
2.2水平集方法6
2.3变分原理8
3主动轮廓模型的介绍10
3.1snake模型10
3.2GVF-Snake模型11
3.3基于水平集方法的图像分割技术12
3.2.1LSWR模型12
3.2.2LSWE模型12
3.2.3LBF模型14
3.2.4SLGS模型16
3.2.5改进的SLGS模型18
4基于局部熵和局部标准差联合驱动的主动轮廓模型20
4.1红外图像的分割现状20
4.2相关理论21
4.2.1图像的局部熵21
4.2.2图像的局部标准差21
4.2.3余弦相似性度量22
4.3改进的用于分割红外图像的主动轮廓模型23
4.4实验结果与分析25
4.5关于如何评价图像分割结果的探究30
结论32
致谢33
参考文献34
1 绪论
1.1 图像分割技术的现状以及传统图像分割方法简介
1.1.1 红外图像的获得及特点如我们所了解的那样，红外图像是指利用红外热成像系统非接触性的测量红外辐射量，将人眼所无法觉察的形象转化为切实可见的清晰图像。过程主要包含两个部分[1][2]：1）首先，需要利用红外探测器将红外辐射转化为电信号。2）电信号通过电视显示系统，经过电子学处理，显示在监视器上，得到红外图像。相比较而言，红外图像普遍地具有以下几个特征[5]：1）弱边缘。普遍性的，红外图像中的目标具有弱边缘，主要是由于图像的获得是通过热成像原理，因此就会导致红外图像中具有很多的干扰，图像背景与人们感兴趣的目标之间的温度界限不明显，就会形成弱边缘。
下一篇：部分相干光束轨道角动量的调控