ACM Reference Format
Chu, H., Chang, C., Lee, R., Mitra, N. 2013. Halftone QR Codes. ACM Trans. Graph. 32, 6, Article 217
(November 2013), 8 pages. DOI = 10.1145/2508363.2508408 http://doi.acm.org/10.1145/2508363.2508408.
Copyright Notice
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted
without fee provided that copies are not made or distributed for profi t or commercial advantage and that
copies bear this notice and the full citation on the fi rst page. Copyrights for components of this work owned
by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish,
to post on servers or to redistribute to lists, requires prior specifi c permission and/or a fee. Request permis-
sions from permissions@acm.org.
Copyright © ACM 0730-0301/13/11-ART217 $15.00.
DOI: http://doi.acm.org/10.1145/2508363.2508408
Halftone QR Codes
Hung-Kuo Chu
1
Chia-Sheng Chang
1
Ruen-Rone Lee
1
Niloy J. Mitra
2
1
National Tsing Hua University
2
University College London
Figure 1: Three halftone QR codes generated by our method. By using a new representation model that minimally binds to the appearance
of QR code, our approach is able to combine halftone images with ordinary QR codes without compromising its readability.
Abstract
QR code is a popular form of barcode pattern that is ubiquitously
used to tag information to products or for linking advertisements.
While, on one hand, it is essential to keep the patterns machine-
readable; on the other hand, even small changes to the patterns can
easily render them unreadable. Hence, in absence of any compu-
tational support, such QR codes appear as random collections of
black/white modules, and are often visually unpleasant. We pro-
pose an approach to produce high quality visual QR codes, which
we call halftone QR codes, that are still machine-readable. First, we
build a pattern readability function wherein we learn a probability
distribution of what modules can be replaced by which other mod-
ules. Then, given a text tag, we express the input image in terms
of the learned dictionary to encode the source text. We demonstrate
that our approach produces high quality results on a range of inputs
and under different distortion effects.
CR Categories: I.4.9 [Image Processing and Computer Vision]:
Applications;
Keywords: Non-Photorealistic Rendering, Halftone, QR code
Links: DL PDF WEB VIDEO CODE
1 Introduction
Quick Response Code, abbreviated as QR code
R
, is a two-
dimensional matrix encoding consisting of black and white squares,
called modules, forming a machine-readable barcode to tag infor-
mation onto products. Originally designed by Denso Wave for the
automotive industry, QR code has quickly been adapted as a fast
and effective way to embed digital content and is extensively used
in diverse fields including manufacturing, marketing, etc. While be-
ing an excellent machine readable format, visually QR code re-
mains a clutter of black and white squares that can easily disrupt
the aesthetic appeal of its parent product.
Since QR codes often take up a non-negligible display area, there
is a growing demand for producing visually appealing QR codes.
Such codes that incorporate high-level visual features such as col-
ors, letters, illustrations, or logos are referred to as visual QR codes.
However, creating a visually interesting QR code without compro-
mising its readability is non-trivial. The key challenge arises due
to the lack of proper understanding or analytical formulations cap-
turing the stability (i.e., validity) of QR codes under variations
in lighting, camera specifications, and even perturbations to the
QR codes [DENSO WAVE 2003; Winter 2011]. Patented and ill-
documented algorithms employed for reading QR codes cause fur-
ther difficulties. Consequently, existing approaches are mostly ad
hoc and often end up favoring readability at the cost of sacrificing
visual quality.
A common strategy to generate visual QR codes relies on inbuilt
error correcting capabilities of QR codes to restore from missing
or corrupted modules (see Figure 2(a)). In absence of suitable an-
alytical or computational support, such approaches involve tedious
trial-and-error runs to produce visual QR codes with little or no
control over the final quality. As a result, the resolution and quality
of the results are strongly dependent on and restricted by the set-
tings used to generate the QR codes. Another heuristic is to mod-
ify a module’s appearance while keeping its concentric region un-
touched, and uniformly blending the neighboring regions with the
code modules (see Figure 2(c)). However, due to the tight bind-
ing to the appearance of QR code, such blending-based approaches
ACM Transactions on Graphics, Vol. 32, No. 6, Article 217, Publication Date: November 2013