Research On Integral Imaging And Light Field Imaging Based On Microlens Array

Recent years, three dimensional imaging and display techniques attract more and more attention. Due to it provides the continuous viewpoints, two-direction disparity and does not require the use of special glasses and special illumination; integral imaging becomes the most promising autostereoscopic display. This dissertation centers on the study of integral imaging and light field imaging based on microlens array. By analyzing and utilizing the disparity in elemental image captured microlens array, the improved depth extraction by linear interpolation, computational integral imaging without intensity non-uniformity and multi-view point reconstruction are proposed and realized. Besides, the accurate simulation of point spread function in light field camera is realized, which set a basis for performance enhancement of light field camera.The main researches and conclusions are listed as follows.1. The extracted depths of object points are assessed by two defined evaluation measurements, the depth precision and the depth correctness. The evaluation results prove the enhancement of depth extraction based on integral imaging by linear interpolation method from two aspects, the whole distribution and the whole deviation. The low resolution of elemental image always limits the application of integral imaging. Without any additional hardware and movement, the correctness and accuracy are increased 15% and 20% respectively, when the elemental images are scaled one time by linear interpolation. In integral imaging, the reconstruction of three-dimensional scene and the extraction of depth can be completed at the same time. It is the advantage that other methods of depth extraction do not have. The quality of reconstructed image becomes better by using the extracted depth information after interpolation. Depth extraction based on integral imaging can be used in the background removal of three dimensional scenes.2. A novel computational integral imaging reconstruction (CIIR) method to remove the grid noise in the reconstructed image of 3D scene is proposed. The grid noise is due to the non-uniform superposition of elemental images (EIs). The intensity of the reconstructed image is the weighted average of the intensity of the inversely mapped EIs. In conventional CIIR methods, the forms of weight have limited effects on the removal of grid noise and lead the lengthy computation. A Gaussian weight is adopted in this paper. That is the edge region of an EI with more number of overlapping has smaller weight than other regions, so that the non-uniform superposition is compensated at arbitrary reconstruction depth. Moreover, the Gaussian weight enables us to shorten the lengthy weight in the conventional CIIR. The improved reconstructed image is helpful to achieve 3D visualization and recognition using integral imaging.3. According to the pin-hole array model and ray optics theory, the multi-viewing point reconstruction method in integral imaging is proposed by imitating the viewing method of human eye. In this method, the information missed in the center elemental image are extracted from the neighboring elemental images and are connected together to form the reconstructed image of three dimensional scene from a certain viewing point. Therefore being different from the traditional method, the view reconstructed in this method is not limited by the number of elemental lens.4. The principle, configure and point spread function of light field camera based on microlens array are studied. The research proves that the two characteristic parameters of point spread function, diameter of spot and the center position of spot affect the quality of reconstructed image and accuracy of three-dimensional information of light field camera. The accurate point spread function simulation are proposed and completed. The proposed numerical condition guarantees the simulated result matches the theories.The main innovations in this dissertation is listed as follows,1. Propose a computational integral imaging reconstruction method with Gaussian weight. The computational integral imaging reconstruction without grid noise is completed. The concise form of Gaussian weight improves the computational efficiency 4%.2. Propose and complete the improved depth extraction in integral imaging with linear interpolation. Define two evaluation measurements. The enhancement by linear interpolation is proved from the two aspects, the distribution and the deviation. The accuracy and correctness of depth extraction are increased by 20% and 15%.3. Propose and complete a multi-viewing point computational integral imaging reconstruction method. Being different from the view of generated by the traditional method, the view generated by this method is not limited by the number of elemental lens.4. Propose the numerical conditions of accurate point spread function simulation of light field camera. The accurate point spread function simulation of light field camera is completed, which set the basis of the improvement of light field camera.