摘要
针对传统分割方法存在分割后结构相似度低、峰值信噪比低和分割效果差的问题。提出考虑渐进有效估计的Probit模型阈值分割算法。对图像的光度和几何信息进行聚类处理,根据纹理特征和边缘特征划分图像,根据图像划分结果和图像的噪声水平对字典进行训练获得稀疏表示先验,结合非局部相似先验建立变分模型,采用迭代重新加权算法求解模型,消除图像中存在的噪声。在混合差分背景分割原理的基础上对图像做分割和匹配处理,通过角点检测方法渲染图像纹理,对图像进行帧分解,提取图像的像素特征。考虑渐进有效估计,构建基于Probit模型的图像阈值分类器,将获取的像素特征输入分类器中,完成图像分割。仿真结果表明,所提方法的图像结构相似度高、峰值信噪比高、分割效果好。
Traditional segmentation methods have the problems of low structure similarity,low peak signal-to-noise ratio and poor segmentation effect.A threshold segmentation algorithm based on Probit model is proposed.Firstly,the photometric and geometric information of the image was clustered.The image was divided according to the texture and edge features.The dictionary was trained according to the image segmentation results and the noise level of the image to obtain the sparse representation prior.Combined with the nonlocal similarity prior,the variational model was established,and the iterative reweighting algorithm was adopted.The model was solved to eliminate the noise in the image.Based on the principle of hybrid differential background segmentation,the image was segmented and matched,the image texture was rendered by the corner detection method,the image was decomposed into frames,and the pixel features of the image were extracted.Considering the progressive effective estimation,an image threshold classifier based on Probit model was constructed,and the obtained pixel features were input into the classifier to complete the image segmentation.The simulation results show that the proposed method has high SNR and good segmentation effect.
作者
王俊彦
李凤萍
WANG Jun-yan;LI Feng-ping(College of Humanities and Information,Changchun University of Technology,Changchun Jilin 130122,China)
出处
《计算机仿真》
北大核心
2023年第2期261-264,共4页
Computer Simulation
基金
“互联网+”背景下民办高校高等数学课程思政教学改革研究与实践
基金号:JGJX2021D493。
关键词
渐进有效估计
图像去噪
像素特征提取
图像分割
Asymptotically effective estimation
Image Denoising
Pixel feature extraction
Image segmentation
