Depth Estimation Method Based On The Research Of Light Field Camera

The light field image is different from the traditional two-dimensional imaging.It contains rich depth information,and the estimation of the depth information can be applied to the corresponding field such as scene reconstruction,depth calculation and multi-angle scenes dyeing.The light field camera can realize the function of shooting first and then refocusing.We can use the scene reconstruction technology to improve the image post processing ability.The extraction of depth information from light field imaging is an important area of current research,and the extraction of depth information has been widely applied to the study of scene depth.More and more domestic and foreign scholars have begun to pay attention to light field imaging technology,and light field imaging theory has been slowly applied to many fields,such as remote sensing,military investigation,microscopic detection,3D reconstruction and stereo vision.The main research contents of this paper are as follows:(6)Firstly,the thesis studies the development process of light field camera and the current research status of depth estimation method.(7)It discusses the concept of the light field,and the method of light field presentation,principle and data acquisition.The key techniques of data extraction and decoding of the original light field are presented as well.The acquired image is decoded,and the four-dimensional light field information of the image is obtained.Besides,a clear view of each perspective is obtained.(8)The thesis researches on digital zoom and digital focus.It also focuses on the image processing algorithm of Stanford-type light field camera.(9)In-depth research on the method and theory of depth information extraction,research on the depth estimation method of polar plane image,multi-view image and refocus focus stack are studied in this paper.It mainly focuses on refocus multi-cue fusion method and image processing applications.(10)This paper mainly studies the image processing method of refocusing multi-cue fusion.Firstly,the consistency depth cues and defocus depth cues of the image are extracted.The confidence map information is optimized by MRF(Markov random field),and the initial depth estimation map of the image is obtained.Then the initial depth map is optimized and globally obtained.After optimizing the depth map,this thesis compares with the results of the occlusion clue optimization depth map.In the case of ensuring the calculation accuracy,it is no longer necessary to calculate the local occlusion cues of the image,which reduces the complexity of the algorithm.