Due to the popularity of smartphones, cameras can be seen everywhere. QR codes are widely used daily, and their application is becoming more and more diverse, such as for warehouse management, electronic tickets, mobile payment, etc. As COVID-19 rapidly spread worldwide, people were forced to change their payment habits. Contactless systems, such as electronic tickets, became increasingly used to display information and avoid traditional queues. However, the standard QR code comprises black and white squares in monochrome images, which is not visually appealing. Yet, the easiest way to present a theme in a QR code is an image, which is more eye-catching and easier to understand than text. In this study, we devise an IS-QR method to integrate full-color images with QR codes by instance segmentation, using BlendMask to extract image feature regions and take Human Visual System into account. Discrete wavelet transform and contrast sensitivity were used to lessen the impact of reduced readability of QR codes during printing. Representative image visual quality measures, including PSNR, MSE, SSIM, FSIM, and GMSD, were used to measure the experimental results in order to validate the effectiveness of QR code beautification. The subjective quality evaluation is also performed. Finally, the measurement results indicate that the beautified QR codes generated by the method IS-QR designed in this study perform better than other related studies in terms of visualization and beautification.
- Deep learning
- QR code beautification