摘要
为了实现对多目标图像和灰度不均匀图像的分割,文中提出了基于区域生长的局部脉冲耦合神经网络(RG-LPCNN)图像分割方法。首先,利用显著性检测方法提取出原始图像的显著性图。然后,根据直方图阈值法对显著性图进行粗分割,得出目标与背景,并将目标的质心作为RG-LPCNN的初始种子点。其次,将高斯核与原始图像的卷积结果作为放大系数,使得动态阈值具有了局部特性。最后,利用RG-LPCNN对图像进行分割,实现对多目标图像以及灰度不均匀图像的分割。将RG-LPCNN和其他阈值分割方法在自然图像、灰度不均匀图像上进行了对比,结果表明:RG-LPCNN在分割多目标图像和灰度不均匀图像方面具有较好的分割效果,验证了RG-LPCNN的有效性。
In order to implement segmentation of images with multi-object and images with intensity inhomogeneity,this paper proposed an image segmentation method based on region growing with local coupled neural networks(RG-LPCNN).Firstly,the saliency map of the original image is extracted by using saliency detection algorithm.Secondly,the object and the background of the saliency map are coarsely segmented by histogram thresholding method,and centroid of the object is taken as the initial seed point of RG-LPCNN.In addition,convolution results of Gauss kernel and original image are used as amplification coefficients to make the dynamic threshold have local characteristics.Finally,the proposed method is utilized to segment images,implementing the segmentation of the images with multi-object and the images with intensity inhomogeneity.The RG-LPCNN algorithm is compared with other thresholding segmentation algorithms in natural images and images with intensity inhomogeneity.The results demonstrate that the proposed method has superior segmentation effect for segmentation of the images with multi-object and the images with intensity inhomogeneity.
作者
王燕
许宪法
WANG Yan;XU Xian -fa(College of Computer and Communication,Lanzhou University of Technology,Lanzhou 730050,China)
出处
《计算机科学》
CSCD
北大核心
2019年第7期258-262,共5页
Computer Science
关键词
显著性
多目标
灰度不均匀
局部特性
Saliency
Multi-object
Intensity inhomogeneity
Local characteristics
