Research On Light Field Refocusing Technology Using Bokeh Rendering

In recent years,the light field imaging in the Synthetic Aperture Imaging(SAI)system has received extensive attention from researchers.Unlike a traditional single camera that can only record the two-dimensional plane information of the scene light,the light field imaging not only records the two-dimensional plane information of the scene,but also records the angle information of the scene light.As a result,images with different focal lengths and depths of field(Do F)can be obtained by setting different parameters,which brings great convenience in image processing.Using these characteristics of computational imaging also brings new processing methods to the fields of parallax estimation,three-dimensional scene rendering,refocusing rendering,image super-resolution processing and other fields.The photos taken by a large-aperture camera have the characteristics of shallow depth of field and high resolution.However,large-aperture cameras are difficult to manufacture and expensive.Synthetic aperture light field imaging uses a camera array comtsed of inexpensive cameras.Sampling of scene light at different angles.The focal length and depth of field can be changed arbitrarily by refocusing the light field.However,the existing light field refocusing methods either face aliasing problems or face a huge amount of calculation.The paper proposes a novel light field refocusing method to improve the image quality based on the light field refocusing of the camera array.This method first obtains scene depth information through disparity estimation,and then renders the defocused area through an anisotropic filter based on a Gaussian distribution model.The main contributions of the paper are:(1)Improve the bokeh rendering model based on the camera array.In the existing bokeh rendering methods based on camera arrays,the light intensity distribution model is uniform distribution.The paper starts with Gaussian distribution to study the bokeh rendering problem.In addition,the existing light field refocusing methods either face the aliasing problem or face a huge amount of calculation.In this paper,bokeh rendering is used to achieve digital refocusing,which solves the aliasing problem within an acceptable execution time.To improve the quality of rendered images;(2)Propose and optimize a bokeh rendering model considering the front and back scenes.Based on the in-depth analysis of the effect of foreground and background on bokeh rendering,a novel technical solution is proposed.It not only avoids the intrusion of the background color information on the foreground objects,but also realizes that the edges of objects at different depths can be well integrated.The thesis will verify the above two points through theoretical analysis and experiments.By comparing with existing methods in terms of time and imaging performance,the method in the paper not only performs well in terms of execution time,but also has higher quality imaging results.