Research On Spectral Light Field Depth Estimation Algorithm Based On Fusion Of Angle Entropy And Adaptive Defocus Information

Spectral information and depth information are two important information describing the scene through the image.However,the traditional imaging method cannot acquire the spectral information and depth information of the scene at the same time,and the accurate description of the scene cannot be realized.In order to better describe the scene through images,this paper combines the theory of light field imaging with the theory of spectral imaging to build a spectral light field imaging system.By collecting and calculating the spectral light field image,the spectral information and depth information of the scene can be acquired simultaneously.However,since the acquired depth information of the single-band spectral light field image is incomplete,which results the acquired depth information is inaccurate.Therefore,the 33 single-band spectral light field images collected need to be combined into a spectral light field true color image.In order to improve the accuracy of the existing depth estimation algorithm,based on the existing depth estimation algorithm,a depth estimation method based on the fusion angle entropy and adaptive defocus information is proposed to realize the depth estimation of the true color image of the spectral light field in this paper.At the same time,considering the large amount of spectral light field image data,which will cause difficulties in the collection,transmission and processing of the spectral light field image.In this paper,33 single-band spectral light fields are reconstructed by the theory of compression sensing,and the true color synthesis and depth estimation processing of the reconstructed 33 single-band spectral light field images are realized,and the depth estimation of the reconstructed spectral light field true color image is realized.The main research contents of this paper are following:(1)The principle and implementation method of spectral light field imaging are studied.Through the research and summary of light field imaging and spectral imaging,a spectral light field imaging system is built.The industrial camera and the liquid crystal tunable filter are combined into a single spectrum acquisition device,and then the three-dimensional translation stage is used to control the object to move at a specific position,and the object of different positions is imaged by a single spectrum acquisition device to collect the spectral light field image of each band.In this paper,33 sets of single-band spectral light field images are collected to form a complete spectral light field image.(2)We study the principle of true color synthesis and color correction of spectral image and its implementation method.By performing true color synthesis and color correction processing on a single spectral image to obtain a spectrally true color image with good quality,and then synthesizing the true color image of the 25 angles of the spectral image in the spectral light field image,the true color image of the spectral light field is obtained.The color image solves the problem that the depth information of the single-band spectral light field image is incomplete.(3)The depth estimation theory and the implementation algorithm are studied.Based on the research and summary of the existing depth estimation algorithm,a fusion angle entropy and adaptive defocusing information depth estimation method is proposed to improve the accuracy of the existing depth estimation algorithm.Firstly,the refocusing is realized by moving and summing the angular patch,then the angular entropy information of the angular patch is calculated to obtain the image consistency information.And the adaptive defocusing information of the image is calculated by using the adaptive defocusing algorithm,then angular entropy information and the adaptive defocus information are combined to obtain the initial depth estimation result.Finally,the initial depth confidence is used as the a priori information,and the initial depth is optimized by the fusion of the scene edge information to obtain the final depth estimation result.A depth estimation of the open light field data set and the original spectral light field true color image is achieved.(4)We study the principle of compression sensing and the implementation method of compressed sensing reconstructed spectral light field.The experimental process of reconstructing spectral light field from compression sensing to depth estimation is completed.Firstly,through the research on the acquisition method of sparse basis,the over-complete dictionary needed for compressing the sensor to reconstruct the spectral light field is obtained,and then the measurement matrix isoptimized.Finally,33 single-band spectra are reconstructed by compression sensing algorithm.The light field image and the true color synthesis and depth estimation processing of the reconstructed 33 single-band spectral light fields is realized.The feasibility of the entire process of the spectral light field from compressed sampling to depth estimation is verified.