Multi-View Data Processing Method For Light Field Display

Humans perceive and recognize the world through vision.In human's daily life,most of the information comes from visual input.As an important part of information transmission,display system is ubiquitous in modern human life.It presents information to viewers in the most intuitive form,which is of great significance.Compared with the conventional display device,the three-dimensional display increases the depth information of the third dimension,and restores the real scene to the greatest extent.The three-dimensional display pursues the restoration of the real-world visual sense as much as possible,prompting the development of 3D display technology toward multi-view,high-resolution,and large-angle perspectives.In order to achieve the desired display effect,3D display requires rich input image information,and its display content has changed from the traditional two disparity maps to the dense light field image,leading to sharply increasing of the amount of data and the tremendous changes of the data structure.This transformation puts higher demands on the collection and processing of three-dimensional information.In practical applications,due to various conditions,it is not possible to rely solely on multi-camera shooting to obtain multi-viewpoint information.When the number of viewpoints is too high,it is necessary to combine light field information processing technology to achieve required information.Based on the above development,this paper proposes a series of multi-viewpoint content processing methods for light field display to realize a variety of three-dimensional display applications.Most of the image processing algorithms currently used for stereoscopic display are still based on binocular stereoscopic display.However,these algorithms are enough for full parallax,multi-view autostereoscopic display.For a large amount of light field data,it is necessary to maintain the consistency of all viewpoint images during processing and to improve the efficiency of the algorithm to achieve real-time processing of light field information.Aiming at the above problems,this thesis studies the multi-view content processing technology of light field display.The main research contents and innovations of the thesis are as follows:(1)dense viewpoint generation method based on lightfieldIn order to more realistically restore the captured real scene content,the three-dimensional display requires a dense viewpoint to improve the visual smoothness of the displayed content.This paper proposes a dense viewpoint generation method based on light field image refocusing.The refocusing algorithm is designed according to the refocusing model,and the images are regenerated to be refocused at different depths.The corresponding depth values can be inferred from these images.Using depth information,a new viewpoint map is drawn through the mapping relationship of pixels.We also propose a method based on interpolation algorithm to generate dense viewpoints.First,the view array is re-encoded to obtain the original light field image according to the characteristics of the light field image and the encoding principle.Then the image is interpolated and the sub-image is extracted from the interpolated image to obtain a dense viewpoint.(2)Free-view walkthrough for the light filed displayTechniques have been proposed for generating virtual views between cameras,using depth information and feature matching between multiple images.However,these methods cannot generate views in front of the cameras and behind the cameras to implement free-view walkthrough for the light filed display.Here a simple and robust method is presented to synthetize virtual views.The key to this technique lies in interpreting the input images as a 4D light filed,and new views are generated by combining and resampling the pre-acquired images.Experimental results show that the method can obtain virtual synthetic views with good visual effects,and the content of the original views can be rightly reflected.