Light-field Super-resolution Via Convolutional Neural Network

Light-field camera is a new imaging system based on the light-field imaging technique.A light-field camera is able to acquire 2D spatial information and 2D angular information of a scene in a single shot by adding a micro-lens array between the main lens and the sensor.As light-field imaging can provide richer information than information provided by a conventional 2D image,light-field has become a hot topic in the image processing field.Aiming at the problem of low spatial and angular resolution of light field camera and combining with the advantages of Convolutional Neural Network(CNN)in image processing,this thesis is focused on super-resolution algorithm of light-field image based on CNN.The main work of this thesis is listed as follows:(1)We summarize the research situation of light-field camera,CNN and super-resolution reconstruction technique,and introduce the basic theory of light-field imaging,the acquisition of light-field,the theory of light-field rendering and the relative theory of super-resolution reconstruction technique and CNN.(2)With multi-view light-field images as input,we design and implement a CNN-based framework to excavate useful information from sub-aperture images to reconstruct every sub-aperture image.Experimental results demonstrate that the proposed method can restore more high-frequency details compared to state-of-the-art methods.(3)We propose a CNN-based framework to augment both the spatial and angular resolutions of light-field images.First,sub-aperture images are inputted to the spatial super-resolution network to restore the high frequency details.Then,reconstructed sub-aperture images are used to generate a novel view by angular super-resolution network.Experimental results demonstrate that in terms of visual effects and evaluation metrics,the reconstruction results of the proposed method is superior to those of state-of-the-art methods.