Optimal embedding of QR codes into color, gray scale and binary images
Date
2014
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
The grow of smart phone and mobile devices market, has created a new set of opportunities for companies to develop new publicity strategies. In particular the association of printed materials with online content is increasingly used. One of the most widespread forms of engaging mobile users from printed materials is based on the use of QR codes, which have been adopted for many different applications such as accessing web sites or downloading premium content. QR Codes are a very reliable and convenient way to introduce textual information into mobile devices without the hassle of typing complicated chains of characters. These applications are outside of the original functional purpose for which these codes were designed in the auto part industry and other considerations besides robustness and speed of decoding have become increasingly important. The problem related to QR code integration into billboards and printed materials presents and important and interesting design challenge since this integration pursue two conflicting objectives: the improvement of visual appearance and the maximization of decoding robustness. This thesis focuses on the development of algorithmic techniques for embedding QR codes into logos or images in order to make them visually appealing to the user while maintaining acceptable decoding robustness. In contrast to previous approaches the methods presented here allows to automatically embed QR codes into color, gray scale or binary images with bounded probability of detection error and minimal intervention of the user. These embeddings are designed to be compatible with standard decoding applications and can be applied to any color or gray scale image with full area coverage. The embedding problem is solved by the integration of different halftoning and visual quality assessment techniques with numerical optimizations specially tailored for each problem. A model of the probability of error at the QR detector is developed and then used in the optimization to yield the best possible combination of transformation parameters for each particular image.