Data Processing And Calibration Method For The Light-Field Cameras

Limited by integral imaging, the captured image of traditional cameras is just a two-dimensional projection of high dimensional visual information of actual scene. To overcome the shortcomings of traditional cameras, a microlens array is inserted between the main lens and the sensor in light-field camera, with the aim to distinguish and record the lights of each pixel from different directions, which can extended the image from traditional two-dimensional to four-dimensional, and thus capture higher dimensional visual information. The main advantage of light-field camera is its ability of refocusing after the fact. The light field camera also addresses the problem that the depth of field is limited by the aperture size. And base on the captured light field image data, the true form of real-time scene can be calculated. Compared with the two-dimensional digital images, light-field image carriers richer information and has a wider f applications.This paper carried out a research around the key technology for applications of Lytro camera, the main contribution is to study and implement the extraction, decoding and preprocessing of light-field data, and the calibration of Lytro camera. The details of this paper are as follows:The first chapter reviews the backgrounds, development history and significance of photography and cameras, analyses the basic architecture and imaging principle of traditional camera, and make a comparison between the light-field camera and traditional camera to reflect the advantages of light-field camera. At last, we listed the main contents of this thesis and gave the relationship of all chapters.The second chapter introduces the definition and acquisition of the light-field. We analysis and study the imaging principle of light-field camera. By inserting a microlens array between the main lens and the sensor in camera, we can capture the four-dimensional light field with the position and orientation information.The third chapter studied and implemented the extraction, decoding, and preprocessing of light-field data. For the white image captured by Lytro camera, we do the following processes, i.e. filtering, Delaunary triangulation,color correction, conversion of perpendicular from hexagons, and extracting the sub-aperture image for camera calibration.The fourth chapter studied and implemented a camera calibration method based on the parameters of dual-plane, that is a series of coordinate transformations among the world coordinate system, the camera coordinate system, the image of the physical coordinates and the image pixel coordinates after extracting the corner from sub-aperture image of original image.The fifth chapter studied and analyzed another method of camera calibration based on a linear feature extraction. Compared with the method introduced in chapter4, it extracts the linear features from original image instead of sub-aperture image, and conducts linear calculation for each pixel, thereby performs the calibration of camera.The sixth chapter summarized the contents of the previous chapters, and forcasted the future development direction and prospects considering the inadequacies of the current light-field camera.