PCA显著性检验结合相关分析的图像缩放法  

A Image Scaling Method of PCA Saliency Detection Combined with Correlation Analysis

作　　者：胡明颖[1] 陶胜[1] HU Mingying;TAO Sheng(Science School,Jimei University,Xiamen 361021,China)

机构地区：[1]集美大学理学院,福建厦门361021

出　　处：《集美大学学报（自然科学版）》2022年第4期379-384,共6页Journal of Jimei University：Natural Science

基　　金：福建省教育厅基金项目(JAT210231,JT180263)。

摘　　要：针对内容感知图像缩放法的显著图,引入主成分分析(PCA)法检测图像的显著性,并结合相关分析法进行图像缩放。先根据图像每个像素点构造3×3领域,通过PCA算法得到每点的显著得分并定义行、列的显著度;再结合图像行列相关分析得到的行列相近度,给出各行各列的重要值,删除或放大较小重要值的行列实现图像的缩放。实验结果表明,该方法理论简单、运行高效,不仅能够完整地保护重要区域,同时还可以让图像的整体概貌过渡良好。A new method of saliency detection based on principal component analysis(PCA)and correlation analysis was proposed for the significance graph of content-aware image scaling.According to the 3×3 fields constructed by each pixel of the image,PCA algorithm is used to obtain the significant score for each point and define the salience of rows and columns.In combination with the similarity obtained from correlation analysis of image rows and columns,the important values of each row and column are given,and the rows and columns with small important values are deleted or enlarged to realize image scaling.The experimental results show that the method is simple in theory and efficient in operation,which not only protects the integrity of important areas but also provides a good transition overall.

关 键 词：图像缩放 显著图 PCA算法 相关分析 重要值 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]