rafaelmaeuer / multiframesuperresolution Goto Github PK

A Multi Frame Super Resolution Tool for Matlab inspired by the "Robust And Fast Super Resolution" tool by Oded Hanson Link. The App was modified and rebuild with Matlab App-Designer, problems with the "Robust Fast Super Resolution" and the "Lucas-Kannade Affine Optical-Flow" algorithms were fixed. The "MATLAB Image Registration" and "Adaptive Kernel Regression" were added as additional options to compute super resolution.

This project was realized in the context of a semester project on the topic "Extended studies of Audio and Video Technologies" (AVT) at the TH Köln.

MFSR Tool to compute an image of higher resolution from a video of low resolution images. Choose from multiple Image-Registration methods and Super-Resolution algorithms. Supported video input formats are AVI, MOV, MP4 and M4V.

Image-Registration methods:

• MATLAB Image Registration
• Lukas-Kanade Optical Flow Affine
• Lucas-Kanade Optical Flow Motion

Super-Resolution Algorithms:

• Adaptive Kernel Regression
• Cubic Spline Interpolation
• Robust Super Resolution
• Fast Robust Super Resolution

The source code is located in MFSR_App folder. The App is build with the Matlab App-Designer and compiled with Matlab Application Compiler.

Compiled standalone apps for Windows and OS X are located in MFSR_Tool folder. If Matlab (R2019b) isn't installed, use MFSR_Installer_web to download the Matlab runtime from web.

Check Video_Samples folder to get some LR video samples to test the different algorithms. Most of them are from

• MDSP Super-Resolution And Demosaicing Datasets Link

Check the Papers folder to get more information about the implemented algorithms.

• Robust And Fast Super Resolution (by Oded Hanson) Link
• Video super-resolution by adaptive kernel regression (by Mohammad Moinul Islam) Link
• Intensity-Based Automatic Image Registration (MathWorks) Link
• Affine optic flow (by David Young) Link

• Fast and Robust Multiframe Super Resolution (by Sina Farsiu et al.) Link
• Pyramidal Implementation of the Lucas Kanade Feature Tracker (by Jean-Yves Bouguet) Link
• Video Super-Resolution by Adaptive Kernel Regression (by Mohammad Moinul Islam et al.) link

• Improve performance (check parallel processing of algorithms)
• Add more configuration parameters in "Advanced Settings" section
• Add option to play LR-Video instead of next and prev buttons
• Optimize GUI adaptiveness when changing window-size or in fullscreen
• Add option to show registered frames in LR-Video preview