Bivariate Empirical Mode Decomposition For Image Retrieval

At present,with the popularization of color cameras and smart phones,there are more and more pictures in life.How to quickly and accurately locate the desired picture among the many pictures has become a hot research topic.Traditional text-based image retrieval requires manual annotation of image categories,which is time-consuming and subjective.In this case,Content-based Image Retrieval(CBIR)has gained more and more attention because of its simplicity and efficiency.CBIR technology mainly completes the recognition and retrieval of pictures based on the characteristics of the picture's color,texture,and contour.Among them,the CBIR technology based on contour features is widely used because contour features are more in line with human visual habits of recognizing images.The step of CBIR based on contour features is generally edge extraction,using shape descriptors to describe edges,matching features,and getting retrieval results.This article mainly studies the shape descriptor to describe the edge.Traditional shape descriptors include Fourier descriptors and shape context descriptors.The Fourier descriptor describes the edge of the image very well from the perspective of frequency.However,in feature matching,the Fourier descriptor only extracts features from the frequency domain and there is only one feature vector,so the ability to describe the edge is limited.The shape context descriptor describes the edge of the image in a statistical method.This method has a strong ability to describe the edge,but requires operations such as uniform sampling,"boxing",and statistics.The algorithm is more complicated and the time complexity is relatively high.To solve the above problems,the main work of this thesis is as follows:1.In the field of contour feature-based CBIR,the traditional Fourier descriptor only extracts features from the frequency domain and there is only one feature vector and the complexity of the shape context descriptor algorithm is too high.Bivariate empirical mode Decomposition algorithm(BEMD)is applied to the field of image retrieval to complete image edge extraction,preprocessing,BEMD decomposition,feature extraction and other related work.2.Design experiments to evaluate the performance of the BEMD-based image retrieval algorithm,select reasonable performance evaluation indicators,and finally show that the BEMD-based image retrieval algorithm has good retrieval performance through a comparison with the Fourier descriptor and the shape context descriptor.