Research Of Color Single-pixel Imaging Based On Compressive Sensing

Single-pixel imaging is a new imaging method,and it's based on compressive sensing theory.In single-pixel imaging,a single-pixel detector can get linear measurements of scene light,and image can be reconstructed by reconstruction algorithm.Single-pixel imaging provides a new path for color imaging.At present,the research of color single-pixel imaging mainly focuses on Single Measurement Vector(SMV)model in compressive sensing,that is,images of red,green and blue channels are reconstructed respectively and then combined into color image.In SMV model,the multi-band structure features of color image are not utilized,and the image quality is not satisfactory.Therefore,this paper proposes to apply Multiple Measurement Vectors(MMV)model to color single-pixel imaging to improve image quality by utilizing the correlation between different bands of color image.SMV model in compressive sensing and monochromatic single pixel imaging based on it were introduced first to pave the way for subsequent color single-pixel imaging analysis.Sparseness of image signal,sampling matrix and reconstruction algorithms of SMV model were analyzed respectively in depth.Wavelet-based sparse transform for image,common sampling matrix and typical reconstruction algorithms were introduced.MMV model and color single-pixel imaging were analyzed subsequently.Basic model and reconstruction algorithms of MMV were analyzed.The correlation of wavelet coefficients in different bands of color image was analyzed,and it's consistent with the premise of MMV model,that is,sparse support of each column in the sparse source signal matrix is the same.The basic framework of color single-pixel imaging based on MMV model was proposed.The simulation results show that the color single-pixel imaging based on MMV model is superior to SMV model in terms of subjective visual perception and objective evaluation index.Finally,a color single-pixel imaging experiment platform was set up.Color imaging experiments were performed on different target scenes.Experimental results show that color single-pixel imaging based on MMV model has better actual imaging performance than SMV model.