融入图像全局信息的局部图像拟合模型  

Local image fitting model fused with global image information

作　　者：陈书贞[1] 甄延海[1] 

机构地区：[1]燕山大学信息科学与工程学院,河北秦皇岛066004

出　　处：《光学技术》2013年第5期466-471,共6页Optical Technique

基　　金：河北省自然科学基金(F2010001294)资助项目

摘　　要：为提高局部图像拟合(LIF)模型对初始轮廓的鲁棒性,提出了一种新的融合图像全局和局部信息的活动轮廓模型(LIF_GI模型)。针对灰度均匀图像,利用图像的全局均值构建了全局图像拟合(GIF)模型,结合GIF模型和LIF模型的优势,通过构造新的图像拟合函数构建了LIF_GI模型。为避免对水平集函数进行繁琐的重新初始化操作,使用反应扩散(RD)方法实现水平集演化。实验表明,所构建的GIF模型在灰度均匀图像上能够获得满意的分割结果,且容许灵活的轮廓初始化。在分割灰度不均匀图像时,LIF_GI模型有效地降低了LIF模型对初始轮廓的敏感性,与LIF模型相比,LIF_GI模型又表现出迭代次数少、检测速度快的优势。To improve the robustness of the Local Image Fitting（LIF）model against the initial contour, a new energy model （LIF GI model） combining both global and local image information is proposed. A Global Image Fitting （GIF） model for images with intensity homogeneities is constructed, which use the global averages of image intensity. The LIF _GI model is constructed by defining a new image fitting function, which integrates the advantages of both GIF model and LIF model. To avoid the complex re-initialization procedure, a novel reaction-diffusion （RD） method is applied to level set evolution. Experimental results demonstrate that the GIF model successfully segments the images with intensity hom ogeneities and allows flexible contour initialization. When used to segment images with intensity in-homogeneities, the L1F_GI model alleviates the sensitivity of the LIF model on the initial contour effectively. Compared with the LIF model, LIF_GI model also show the advantages of less iteration and lower time cost.

关 键 词：图像分割 图像拟合函数 局部图像拟合模型 全局图像拟合模型 融入图像全局信息的局部图像拟合模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]