基于特征语义模型的复杂结构图像分割算法  

The semantic feature model on image segmentation with complex structure

作　　者：赵彦会 崔嘉[1] 董新锋[1] 王化雨[1] 郑元杰[1] ZHAO Yanhui;CUI Jia;DONG Xinfeng;WANG Huayu;ZHENG Yuanjie(School of Information Science&-Engineering,Shandong Normal University,Jinan 250014,China)

机构地区：[1]山东师范大学信息科学与工程学院,济南250014

出　　处：《扬州大学学报（自然科学版）》2018年第3期60-64,共5页Journal of Yangzhou University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61502285);山东省自然科学基金资助项目(ZR2014FQ013)

摘　　要：针对结构复杂图像所具有的相似复杂性和属性复杂性,提出了基于特征语义模型的图像分割算法,通过采用矩阵分解将图像分割问题转化为最优化问题.对原始图像进行视窗扫描,选取规模为n的视窗计算所有像素的特征语义.经过PCA降维后,在保证特征信息最小丢失的前提下计算相互之间的特征语义相似度.通过最优化理论近似寻找特征语义相同的像素点集,从而完成结构复杂图像的分割.经过实验证明,本文方法针对结构复杂的图像物体分割完整,细节分割精细.通过对比实验验证了该方法的有效性.In order to solve the image segmentation on images with similarity complexities and property complexities,a novel algorithm according to the semantic hashing model is proposed.The problem of segmentation at pixel level is reformulated in form of the optimization issue.The n scale patch window is used to scan the image for hashing feature vector computation.After data preprocessing by the PCA,the semantic distance between every pixel-pairs can be calculated in an efficient way.By the new similarity computation function,the image can be segmented by the feature points^retrieval.Several experiments are proceeded to verify the feasibility of the algorithm proposed.

关 键 词：图像分割 语义检索 谱哈希算法 矩阵分解 复杂结构 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]