Abstract: A novel image processing method is proposed for extraction of road cracks from fuzzy and discontinuous pavement images. Firstly, a series of pre-processing operations are implemented to enhance the differences between cracks and backgrounds, which include contrast enhancement, de-noising based on Canny-HBT filter and brightness non-uniformity correction; secondly, the fusion of detector responses is performed using multi-scale curvelet transform and maximum-
approximation with mean-detail image fusion algorithm to make threshold decision on the cracks connected area to realize the fine cracks fusion. The experimental results show that the proposed algorithm can make improved edge detection in images with superior edge localization, and can extract the small cracks and the cracks with the weak contrast effectively, and it improves the reliability and precision of the crack detection with higher PSNR.

Key words: pavement crack, preprocessing, Canny-Hyperbolic Tangent function（HBT） filter, curvelet transform, fusion

摘要： 针对模糊、不连续的路面裂缝病害图像，提出一种新颖的图像处理算法进行裂缝病害的提取。将采集的路面病害图像进行对比度增强、Canny-HBT（双曲正切）滤波去噪、亮度非均匀校正等一系列预处理操作，以提高图像中裂缝与路面背景的差异；对预处理后的路面病害图像进行多尺度的曲波变换表示，并在图像重构前利用Max-Mean混合的融合算法对裂缝连通区域进行阈值判决实现裂缝的精细融合；实验表明，该算法能更好地优化边缘定位，有效地提取弱对比度的裂缝和细小裂缝，提高了路面裂缝边缘检测的可靠性和精度，同时也提高了PSNR值。

关键词: 路面裂缝, 预处理, Canny-双曲正切（HBT）滤波器, 曲波变换, 融合