在复杂光照条件下二维码扫码器采集到的图像容易出现整体高亮、阴影区域和局部高亮、阴影区域, 使得图像分割阈值确定困难, 研究了Sauvola算法中的窗口大小w值和修正因子k值对于QR码图像二值化的影响。针对全局二值化方法抗噪能力差和局部二值化方法处理速度慢的缺陷, 提出了一种改进的QR码图像二值化方法, 将Otsu和Sauvola算法相结合提升算法抗噪能力, 并利用积分图算法提高算法运行效率。实验证明, 该方法二值化效果优于经典的二值化方法, 平均运行效率比原Sauvola算法提高17倍, 提升了识别成功率。

In complex illumination conditions, the images collected by two dimensional code scanner are prone to general highlighting with shade and part highlighting with shade, which make it difficult to determine the image segmentation threshold. In Sauvola algorithm,the effect of the window size w value and the correction factor k value on the quick response (QR) code image binarization was studied. Aiming at the shortcomings, such as the poor anti-noise ability of the global binarization method and the slow processing speed of the local binarization method, an improved QR code image binarization method was proposed, which combined the Otsu and Sauvola algorithms to improve the algorithm anti-noise ability ,and used the integral graph to improve the algorithm operation efficiency . Experiments show that the binarization effect of this method is better than the classical binarization method, and the average running efficiency is 17 times higher than that of the original Sauvola algorithm, which increases the recognition success rate.