Theory Of Plenoptic Imaging In Wave Optics And Reconstruction Of Light Field

Light field photography or in other word plenoptic photography is an important branch in the field of computational imaging. It combines the hardware design of the optical system and the software algorithm of the computer. In the aspect of information collection, the system is designed with special structure to acquire numbers of light field information just in one exposure. The data recorded by plenoptic camera include spatial and direction distribution information of the target object. Therefore we consider the plenoptic data as the 4-dementional information. In the aspect of information process, we use several algorithms to realize light field reconstruction which plays dominant roles in many applications and meet the different needs, such as digital refocusing and depth information extraction. This new theory and techniques greatly exceeded the limitations of conventional photography.In the paper, plenoptic photography base on the micro lens array is discussed in three aspects including theory, simulation and information processing. Through the imaging system, the imaging results are obtained from the input image and then the information is reconstructed through the reverse processing, which form a closed loop.Firstly, we model plenoptic imaging system. According to the matching conditions which is derived and with reference to the existing lens and micro lens array technology level, in this paper a set of specific device parameters and structure parameters is given and those parameters provides the necessary data for the subsequent simulation. From the perspective of geometric structure, the formula of depth of field is deduced that Quantify the performance of the system expands the depth of field. Then, the scalar diffraction theory of plenoptic imaging model is established based on the model and parameters which are selected before, we can simulate point spread function with varied object distance. Combined with optical sectioning three-dimensional object imaging model, the forward evolution process of target under three different conditions can be achieved. Based on the light field information obtained from the forward process, several kinds of views can be extracted and retrieved. In this paper Two types of reconstructions are researched. The first one is the re-focusing. the re-focusing of light field in spatial domain is established. The clarity-evaluation-function of gray gradient is proposed to definition the quality of reconstruction image. The second part of research is deconvolution reconstruction.The nearest neighbor domain method is proposed to preliminarily realize the three-dimensional deconvolution. Through the deconvolution the object information of each layer is obtained. The results of deconvolution are also evaluated. As fortranslucent objects which have absorption characteristics, the imaging model and deconvolution model are revised.