摘要
不同图像具有不同的特征和结构,为了更好分析和识别图像信息,提高特征分类精度,提出了基于高斯函数的显著图像特征提取方法。利用指数函数法感知显著图像轮廓边界,将图像变换到含有明暗信息的色彩空间,并锐化算子,判断抑制或增强范围,筛选出显著图像的信息。按照不同的目标函数计算映射矩阵,保留原空间数据点的局部信息,赋予最佳限制条件,提取显著图像的局部特征,结合高斯函数估算协方差矩阵获得高斯特征,完成全局特征提取。通过实验证明,所提方法对三个数据集的识别准确率在85%以上,且特征提取效果好,纹理细节清晰,保证图像信息完整。
Different images have different features and structures.In order to better analyze and recognize image information and improve the accuracy of feature classification,this paper presented a method for salient image feature extraction based on Gaussian function.First of all,the exponential function method was adopted to perceive the con⁃tour boundary of the salient image,and then the image was transformed into the color space containing light and dark information.Meanwhile,operators were sharpened,and the information of the salient image was filtered out by jud⁃ging the inhibition or enhancement range.According to different objective functions,the mapping matrix was calculat⁃ed,and the local information in original spatial data points was retained.Moreover,the best constraints were set to extract the local features of the salient image.Furthermore,the covariance matrix was estimated by the Gaussian func⁃tion.After that,the Gaussian features were obtained.Thus,the global feature extraction was completed.Experiment results prove that the proposed method has a recognition accuracy of over 85%for three datasets,with good feature extraction performance and clear texture details,which ensures the integrity of image information.
作者
胡艳婷
王璐娜
张通
李明亮
HU Yan-ting;WANG Lu-na;ZHANG Tong;LI Ming-liang(Huaxin College,Hebei GEO University,Shijiazhuang Hebei 050700,China;School of Information Engineering,Hebei GEO University,Shijiazhuang Hebei 050031,China)
出处
《计算机仿真》
北大核心
2023年第12期232-235,共4页
Computer Simulation
基金
2021年河北省人力资源与社会保障研究课题(JRS-2021-5028)
2022年河北省人力资源与社会保障研究课题(JRS-2022-3173)。
关键词
高斯函数
图像特征提取
显著图像
黎曼流形
颜色预处理
Gaussian function
Image feature extraction
Salient image
Riemannian manifold
Color preprocess⁃ing
