burden on QR code decoder, which may result in a longer decoding
time or even un-decodable fault. In our experiences, a
normal QR code decoder may not capable of decoding the QR
codes with version size larger than 20. Of course, the decoding
ability depends on the capability of equipped QR code decoder
on the mobile phone.
SA optimization framework is adopted in this work because
the saliency considerations can be seamlessly integrated into
the choice of neighboring states (c.f. (6)). Since SA optimization
has an explicit state transition behavior as compared with
other optimizationmethods (e.g.NeuralNetworks),we can integrate
the saliency consideration into the state transition behavior
easily.We do believe that SA optimization is just one of the possible
solutions to accomplish the required beautification task.
The ease of implementation and the ability to achieve seamless
integration makes SA on the top of the candidate list. Actually,
as long as the saliency considerations can be successfully incorporated
into the optimization procedure, other optimization
approaches, such as Genetic Algorithm and Neural Networks,
may also be utilized in the proposed framework.
Of course, how to find an optimization scheme achieving the
best trade-off among the visual pleasant of the beautified QR
code, the amount of embedded message, and the decoding speed
is the main target of our future work.