High Resolution Remote Sensing Image Retrieval Using Quin-Tree and Multi-Feature Histogram
Yulin Xie · Geo-information Science · 2010
Nowadays,vast amount of remote sensing data have been acquired with the rapid development of Earth Observation System(EOS).It has become a serious task to manage and use these data for most RS and GIS applications.The content-based retrieval system for remote sensing images(CBRSIR) has become resultingly a hot research field with the potential to retrieval interesting information from image databases automatically and intelligently.In this work,we put forward a new remote sensing image retrieval approach by using multi-features including image color and texture.Firstly,a given image is processed by principal components analysis and then decomposed by Quin-tree,which splits large-scale remote sensing imagery into sub images.Secondly,texture features of each image block are extracted via multi-channel Gabor filters,and the standard deviation and third moment of each sub image are extracted as color features.Then,color and texture histograms are constructed based on sub images.Finally,we compare the similarity of the color and texture histograms between the query example and each one in the image database.If the total similarity is higher than some threshold,the image will be returned.These images are sorted according their similarity as the final retrieval results.This approach is validated using high resolution remote sensing images.