Study On Decoding Optimization Algorithm Of Light Field Image Based On Microlens Structure

As a new type of multi-eye imaging device,the light field camera can capture 4D light field information which includs two-dimensional angle information and two-dimensional spatial information by one exposure.The light field image contains more information than traditional 2D image,and depth information in the scene can be obtained from 4D light field image,which plays an important role in the field of machine vision such as virtual reality,3D reconstruction,and target detection.However,the raw lenslet image of the light field has serious problems on specific vignetting phenomenon and edge pixels aliasing of each microlens image.Aiming at these problem,this thesis focuses on the decoding optimization algorithm of light field image,and realization of vignetting correction and demosaicing of light field camera.The main work of this thesis is as follows:(1)We expound the background and research status of vignetting correction method and light field demosaicing method,analyze the research significance of this thesis,and introduce the theory of light field data decoding process,vignetting correction theory and light field demosaicing theory.(2)According to the formation reason and characteristics of the special vignetting phenomenon in raw light field lenslet images,we design a light field vignetting correction method using constrained radial gradient symmetry.Firstly,the global vignetting correction is performed on the whole raw lenslet image by using the white image of the light field camera.On this basis,for the residual vignetting of each microlens image,the radial gradient distribution symmetry of non-vignetting image is constrained by the accuracy of the vignetting function value and vignetting non-increasing characteristic.In this way,the local vignetting correction is realized.The experimental results show that the method can effectively reduce the special vignetting phenomenon caused by the main lens and each microlens in the light field camera,thence each pixel of the light field image has high brightness consistency.(3)Aiming at the problem that the light field multi-view image quality is poor due to pixel aliasing and vignetting effect at the edge of the light field microlens,we design a light field demosaicing algorithm based on double-guided filtering.First,the G image is reconstructed by reweighting the gradient based threshold free(GBTF)algorithm with the white image and lenslet mask information,Then,the reconstructed G image is used to double-guide the R/B image for reconstruction.Finally,the reconstructed R,G,and B images are combined into a full color image.The experimental results show that the method effectively reduces the influence of edge pixel aliasing and vignetting effect on the demosaicing of the light field,and improves the quality of the full-color image of the light field.