Three Dimensional Reconstruction Based On Light Field And Digital Refocusing Technology

There are increasing people using 3D images to express information in more and more areas, thus enabling us to better perceive the diversity of the real world.3D virtual model is able to reproduce 3D landscape, achieve spatial roaming and spatial analyzing, and support decision more intuitively. To restore the original appearance of the scene, it requires the accurate 3D spatial data; it's one of the problems to be solved that to access to such information effectively.In this paper, studied the key technologies of light-field imaging involved, summed up the light-field theory, combined with definition of light-field and light-field rendering theory described the application of light-field in 3D model reconstruction. Furthermore, designed light-field imaging simulation software by micro-lens method and 4D parametric light-field; analyzed the advantages of light-field camera in 3D digital imaging compare to usual camera.Using a method based on light-field and digital refocusing technology to obtain depth information; used the light-field to collect data, calculated the refocusing image sequence, then used the frequency domain or gradient interpretation function to obtain the depth information. Digital refocusing solves the complex problem of the sampling process of DFF. The method was simple and has high precision, which applying to the depth measurement is equivalent to an improved focusing distance measuring method.Devised a method of adaptive image enhancement based on mode collaborative filtering, to improve the depth map from light-field. This method used mode to reflect the general level of data, formed a balance around the mode and isolated deviation value by collaborative filtering. Used it to deal with the depth map with the above unique properties, and leaded to a result that it can get better effect than other commonly filters.Using the depth map from light-field method and mode collaborative filtering to 3D model reconstructed; based on VC++, OpenGL and ArcGIS Engine, designed a reconstruction and browsing tool for the depth map and omni-focus image above. The experimental results show that the method mentioned in this article can achieve good results.