Robust Hazy QR Code Recognition based on Dehazing and Improved Adaptive Thresholding Method
Zhengcai Cao, Junnian Li, Biao Hu · 2019
Quick Response (QR) code has been extensively used in our daily life. But in some complex environments such as hazy conditions, QR code recognition is difficult. In this paper, a robust QR code recognition algorithm in complex hazy environments is proposed. First, contrast-limited adaptive histogram equalization is implemented to enhance original pictures. Then Gated Context Aggregation Network is modified to obtain dehazed QR code images. After that, an adaptive thresholding method is used to obtain binary images. Finally, binary QR code is decoded. For training and testing our algorithms, we collect a set of benchmark of hazy QR code images including hazardous chemical name, Chemical Abstracts Service number, shape and properties such as flammable, explosive, corrosive and toxic. Ablation comparison and experimental results on our own database demonstrate our proposed algorithm achieves superior performance on hazy QR code recognition tasks.