Research On Light Field Reconstruction And Depth Estimation Based On Compressed Sensing

The light field image containing rich spatial 3D information can be used for re-focusing,depth estimation and three-dimensional display.The precise depth information plays an important role in significant detection,super-resolution reconstruction,target recognition and 3D surface reconstruction technology.Although the two methods of camera array and the micro lens array can effectively record the light field image,the former method is limited due to its large size and high cost.The latter method is based on the spatial resolution of the sacrificial image in exchange for the angular resolution.Therefore,in this paper,the mask method is used for high-resolution light field acquisition and reconstruction,to complete the depth estimation based on the reconstructed light field image.The main contents of this paper are as follows:(1)This paper studies the correspondence between the basic principle of compressed sensing and the mask field camera,and applies the compressed sensing theory to the reconstruction of light field.The sparse representation of the signal is the precondition of the reconstruction in compressed sensing theory.The principle of K-SVD algorithm is expatiated in detail.The K-SVD algorithm is used to train the light field sample set to obtain the over complete dictionary of the light field in order,to satisfy the ideal sparse representation of the light field image,and to better reconstruct the light field image.(2)The optimization of random measurement matrix satisfies the need of physical reconstruction.The simulation of the optimized random measurement matrix light field reconstruction.The simulation results show that this method can reconstruct the image of high spatial resolution and large angle resolution.On this basis,real physical platform acquisiting light field can be built depend on compressed light field.Under this platform,transformation relationship among the physical mask,projection matrix and measurement matrix is researched,Completing the transformation from mask to the measurement matrix.Finally,compressed light field physical acquisition and reconstruction can be realized through the light field over-complete dictionary,measurement matrix obtained from mask transform,CCD coded sample image and the compressed sensing reconstruction algorithm.Which provides a simple and effective method for the acquisition of real light field images.(3)Analysis of light field image re-focusing principle,with light field angular patch shift sum replacement complex points to achieve light field image re-focusing.In order to better estimate the depth information of the occlusion edge,this paper firstly analyzes the characteristics of the light field image to optimize the angular patch in the presence of the occluded part.To solve the problem of depth field image occlusion,edge profile information is used to optimize the depth map.Compared with the method of occlusion cue to optimize depth map,this method can reduce the complexity of algorithm and guarantee the accuracy of calculation.Finally,the reconstructed light field image is used to re-focused and depth estimated.