Light Field 3D Reconstruction Method Based on Fast Fourier Transform

Abstract

The invention discloses a method for three-dimensional reconstruction of light field based on fast Fourier transform. The method comprises: converting the collected light field map stack into 4D phase space information; rearranging pixels at corresponding positions in the light field map , converted into multi-angle phase space information; initialize the three-dimensional volume information according to the phase space information, and then convert it into three-dimensional frequency domain information; arrange the point spread functions of the five-dimensional phase space according to two-dimensional angles, and obtain the three-dimensional point spread corresponding to each angle The spatial information of the function transforms the point spread function into the frequency domain; the 3D volume is forwarded to obtain the 2D frequency domain information of the forward image; The Fourier transform is transformed into spatial domain information; the collected phase space information is transferred to the frequency domain, and transformed into spatial domain information through backpropagation and inverse Fourier transform; the spatial information of the three-dimensional volume is updated until the end of the iteration. This method can quickly process light field data and obtain accurate 3D reconstruction results.

Description

technical field [0001] The invention relates to the technical field of three-dimensional reconstruction, in particular to a three-dimensional reconstruction method of light field based on fast Fourier transform. Background technique [0002] Light Field (LF) imaging is a fast volume imaging method. The light field system adds a microlens array to the original imaging surface, and uses sensors to record 4D light field information. Since it can record the 3D information of the scene with a single exposure, it has been widely used in the field of 3D imaging in recent years. [0003] For the 4D light field information collected by the light field system, the existing three-dimensional reconstruction methods include Richard-Lucy reconstruction algorithm, reconstruction method based on phase space, etc., but the use of the above methods often involves a two-dimensional convolution process, and the speed of reconstruction is related to the speed of the image. Size, depth of recon...

AI Technical Summary

Problems solved by technology

[0003] For the 4D light field information collected by the light field system, the existing 3D reconstruction methods include the Richard-Lucy reconstruction algorithm, phase space-based reconstruction methods, etc. The size and the depth of the reconstruction are positively correlated, the speed is very slow and it will take up a lot of computing resources

In addition, using the Richard-Lucy reconstruction algorithm will also bring artifacts to the focal plane

Images