Image Retrieval Method based on Integration of Principal Component Analysis and Multiple Features

Abstract

Jingji Zhao School of Mathematics and Statistics, Nanjing University of Information Science & Technology, \u00a0 Nanjing, 210044, China (Received May 11 2019, accepted July 20 2019) Existing \u00a0content-based \u00a0image \u00a0retrieval \u00a0methods \u00a0exist \u00a0some \u00a0drawbacks, \u00a0such \u00a0as \u00a0low \u00a0retrieval precision, \u00a0unstable \u00a0performance. \u00a0To \u00a0address \u00a0these \u00a0drawbacks, \u00a0in \u00a0this \u00a0paper \u00a0a \u00a0content-based \u00a0image \u00a0retrieval method \u00a0is \u00a0presented \u00a0based \u00a0on \u00a0multi-feature \u00a0fusion \u00a0of \u00a0principal \u00a0component, \u00a0oriented-gradient \u00a0and \u00a0color histogram. \u00a0The \u00a0idea \u00a0for \u00a0the \u00a0proposed \u00a0method \u00a0is: \u00a0firstly, \u00a0input \u00a0image \u00a0is \u00a0grayscale \u00a0and \u00a0flattened \u00a0into \u00a0a \u00a0one- dimensional vector, and the first n principal \u00a0components from the vector yielded by the \u00a0PCA algorithm \u00a0are extracted, in other word, input image is represented as \u00a0a n\u00d71 dimensional PCA \u00a0feature \u00a0vector. Secondly, to remedy color and orientation information missed by PCA, oriented-gradient and color histograms are used to extract orientation and \u00a0color features \u00a0respectively. Thirdly, extracted oriented-gradient \u00a0and color histograms are \u00a0merged \u00a0with \u00a0PCA \u00a0features \u00a0to \u00a0generate \u00a0the \u00a0multi-feature \u00a0representation \u00a0of \u00a0the \u00a0input \u00a0image. \u00a0This \u00a0paper confirms that the proposed multi-feature method can better represent an input image and can easily measure the similarity between images. The experiments are carried out and evaluated based on Corel-1000 , the target method is significantly better than the four popular methods.