Abstract: An improved algorithm of line detection is proposed to aim at the drawbacks of Hough transform. It reduces algorithm complexity based on the principle of ‘many-to-one’. The idea ‘from partial to global’ is also adopted. The parameters of line are gotten in sub-areas in order to reduce quantization error. Then it searches other points of the line within whole image according to the parameters. Algorithm complexity is also analyzed. Experiments show that the algorithm has better properties of computing efficiency and robustness.

Key words: Line detection, Hough transform, Quantization error, Algorithm complexity

摘要： 针对哈夫变换的不足，提出了一种改进的直线检测算法。该算法基于“多对一”的原理，降低了算法复杂度，利用从局部到全局的思想，在局部子区域中获取直线参数以减少量化误差的影响，然后据此参数扩展至全图范围搜索，完成直线检测，分析了算法的时间和空间复杂度。实验表明，该算法占用存储空间小、运算快速、且具有较强的鲁棒性。

关键词: 直线检测, 哈夫变换, 量化误差, 算法复杂度