Reconstruction And Depth Information Extraction Of Light Field

People record a variety of information through the cameras,while the traditional digital cameras can only record two-dimension information,they are lack of perception of the depth information.Light field is becoming a hotspot of many researchers and institutions all over the world.They had made many achievements in the appearing of sheltered objects,layering and separating,refocus and defocus,and the reconstruction of three-dimension,the main technologies for that are light field reconstruction and accurate depth extraction.So,pointing at light field imaging and depth extracting,this paper has carried on a comprehensive research work based on lens-array by using the standard datasets.The main work is as follows:1)This paper proposes a way to reconstruct the light.field captured by lens-array.Firstly,calibrating the center of lens-array images,this paper takes decoding,filtering and color correcting on the image because the unique format of lens-array's light field.Then the altorithm of hexagon-square is used to convert the sub-images to be orthogonal.Lastly,it uses the way of resampling to get the refocus images in frequency domain based on Fourier transform,and transforms the light field to multiple views images for depth extracting by the light field function.2)This paper proposes a algorithm operating on Epipolar Plane Image(EPI)com-pletely.For the dataset captured by lens-array,the traditional algorithms for depth extracting do not work well because of the narrow baseline and disparity range.So,from the EPI aspect,this paper extracts the center view image and horizontal,vertical EPI separately,then uses linear interplolating to get the new pixels in different gradient based on every EPI slice.The confidence level of matching between the two pixels is calculated by a cost function to find the best gradient,then the best disparity map is generated.Finally,it merges and optimizes disparity maps in the arkov Random Field(MRF)to get the best depth map.The experiment results prove that this paper can reconstruct the light field by using the lens-array images,and get the distinct images under different focus and views.Ig-noring the inherent problems in depth extracting of lens-array,the experiment can also create good depth maps.All of the above show that our algorithms is very effectives.