The Light Field Editing repository was created to include scripts of a number of light field editing algorithms. This repository will include code covering different light field editing tools including light field spatial super-resolution, angular super-resolution, light field inpainting and light field denoising, among others. The aim of this repository is to facilitate reproducable research in the area of Light Field image processing which can be used as a benchmark other algorithms.
The LF_super_resolution_analysis script is a command line script that allows to analyse the performance of a number of light field spatial super-resolution algorithms. This script considers that the light field is down-sampled by a scale factor defined by mf and up-scaled to the target resolution using bi-cubic interpolation. The syntax to call this from the command line is:
LF_super_resolution_analysis('bicubic',3,true
Input:
sr_method: specifies the super-resolution algorithm to be simulated. The following is a list of sr_methods that are supported here:
-
'bicubic': classical bicubic interpolation of each sub-aperture image independently
-
'lf_srcnn': this method applies SRCNN to restore each sub-ape[rture image separately from the others [1],[2]
-
'pca_rr': this method applies the pca_rr published in [4] - super-resolves patch volumes
-
'bm_pca_rr': this method applies the pm_pca_rr published in [4] - super-resolves aligned patch volumes.
-
'pb-vdsr': this method is the proposed method using VDSR to restore the principal basis. The code is not yet available.
-
'pb-lab402'; this mehtod is the proposed method using lab402 method to restore the principal basis
mf: numeric value that stands for the magnification factor that the method has to super-resolve
out_flag: This is a boolean value which specifies if the result attined in the simulation will be stored or not. By default it is set to false
[1] Y. Yoon, H. G. Jeon, D. Yoo, J. Y. Lee and I. S. Kweon, "Light-Field Image Super-Resolution Using Convolutional Neural Network," in IEEE Signal Processing Letters, vol. 24, no. 6, pp. 848-852, June 2017.
[2] C. Dong, C. C. Loy, K. He and X. Tang, "Image Super-Resolution Using Deep Convolutional Networks," in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 38, no. 2, pp. 295-307, Feb. 1 2016.
[3] J. Kim, J. K. Lee and K. M. Lee, "Accurate Image Super-Resolution Using Very Deep Convolutional Networks," 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, 2016, pp. 1646-1654.
[4] R.A. Farrugia, C. Galea, C. Guillemot, "Super Resolution of Light Field Images using Linear Subspace Projection of Patch-Volumes," in IEEE Journal on Selected Topics in Signal Processing, vol. 11, no. 7, pp. 1058-1071, Oct. 2017
- You have to install matconvnet and include the matconvnet library into the folder MATLAB/
- Store all the light fields in a directory ../LF-DATASET/. We consider here four different datasets
This system was implemented and tested on Windows 10 OS. or Windows 10 users, we highly recommend installing Bash shell to run Linux commands.