一种新的结合图像梯度的局部活动轮廓模型  

New local active contour model combined image gradient

作　　者：刘驰[1] 孙权森[1] 

机构地区：[1]南京理工大学计算机科学与工程学院,南京210094

出　　处：《计算机应用研究》2014年第11期3510-3513,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(61273251)

摘　　要：基于区域的局部二值拟合模型只考虑图像灰度的平均值统计信息,然而由于图像中的噪声改变了图像的灰度分布,该类方法对于包含大量噪声的图像往往很难获得理想的效果。为了提高模型对于噪声的鲁棒性,提出了一种结合图像统计信息和梯度信息的局部活动轮廓模型。该模型在图像灰度的统计信息的基础上,加入图像梯度信息,分别构造以高斯函数为核函数的局部二值灰度拟合能量和局部二值梯度拟合能量,得到最终的能量泛函,并通过最小化该能量函数,驱使活动轮廓向目标边缘演化。实验结果表明,基于图像灰度和梯度的局部活动轮廓模型能够有效克服图像中弱边缘以及强噪声对于分割结果的影响,其分割精度高于同类方法。Local binary fitting model based on area only considers image statistical information.But the noise in image changes image intensity distribution,so this model cannot efficiently segment images with heavy noise.In order to improve the robustness to the noise,this paper proposed a novel local active contour model driven by image gray and gradient.This model that com-bined the image gradient information is on the basis of the image intensity statistics.The energy function of this model was then constructed with two parts：the local binary gray fitting energy and the local binary gradient fitting energy.Finally,the contour could be driven to the objective boundaries by minimizing the energy functional.Experimental results show that the proposed model can overcome the impact of the weak edges and heavy noise in the image to get satisfied segmentations,and outperforms state-of-the-art algorithms.

关 键 词：C-V模型 局部二值拟合模型 局部灰度 局部梯度 活动轮廓模型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]