Structured Sparse Low Rank Model And Its Application In 3D Image Processing

With the continuous development of image imaging technology,image is important in the fields of satellite remote sensing and clinical medicine.In these applications,the images collected are usually three-dimensional(3D)images.Therefore,the representation and processing of 3D image are very important.Sparse and low rank representation model is widely used in three-dimensional images and has achieved many significant results.However,the existing sparse low rank model is still not perfect for structural feature representation,which is worthy of further study.To further improve the structured sparse low rank model and improve the performance of 3D image representation,this paper studies the structured sparse low rank model of 3D image and its application in 3D image processing.The main research work and innovation of this paper are as follows:1.A triple low rank model is proposed and applied to many kinds of 3D image processing.By deeply analyzing the internal structure between the low rank part and the original matrix in the traditional low rank decomposition model and the autocorrelation constraints of the low rank part,this paper proposes a triple low rank model,which can further improve the robustness and the separation accuracy of the sparse part and the low rank part.The model is then applied to hyperspectral image denoising and medical image denoising.The experimental results show that the triple low rank model is effective in different 3D image processing fields.2.A multi-component low rank dictionary learning model is proposed and applied to a variety of 3D image processing.To effectively represent various components in 3D images,this paper proposes a multi-component low rank dictionary learning model,which uses different dictionaries for sparse representation of different components,and considers the low rank characteristics of the dictionary.A joint low rank dictionary learning algorithm is designed for multi-component dictionaries learning simultanously.In addition,the model is applied to 3D medical image fusion and hyperspectral image denoising.The experimental results verify the effectiveness of the multi-component low rank dictionary learning model.