The UC3M QR database
====================

This database contains several QR codes captured from the screen of several mobile devices using another mobile device.

Database structure
==================

The structure is very simple, containing three folders:

- codes: contains the 20 codes used to construct the database. The name of each file is a concatenation of the code number and the numerical information contained in the code.

- captures: contains 21 captures using different display devices and codes. Each subfolder contains several frames taken from a video of the process of a user focusing the code.  That is reason why initial frames are rarely useful.

- sets: contains 3 files that show the division of the database into 3 sets: train, val and test. Each file shows serveral rows, each of them containing three numbers:
	1.- capture folder
	2.- code associated to the capture
	3.- displaying device used in that capture (see table II in [1]). 


Evaluation
==========

For evaluation, we have only considered those codes that have reached our deblurring module. That means that every code that is discarded in previous stages of our approach [1] is not included in the evaluation.

At the end of our approach, we have used a QR decoder to check if our restoration system successfully recovers the code. You can find the software at:

http://de.sourceforge.jp/projects/qrcode/


Reference
=========

If you are planning to use this database, please include the following reference:

[1] Munoz-Mejias, D.; Gonzalez-Diaz, I.; Diaz-de-Maria, F.; , "A low-complexity pre-processing system for restoring low-quality QR code images," Consumer Electronics, IEEE Transactions on , vol.57, no.3, pp.1320-1328, August 2011.
