Abstract: In Curvature Scale Space（CSS） corner detection algorithm, Gaussian filter is used to remove noise. However, this method reduces the accuracy and precision of the corner detection, therefore, this paper proposes a method that bilateral filter is used for improving the CSS algorithm. It can not only remove noise, but also guarantee the high precision. The large number of experiments fully validate the efficiency and accuracy of the proposed algorithm compared with other algorithms.

Key words: corner detection, Curvature Scale Space（CSS）, bilateral filtering, Gaussian filtering

摘要： 在曲率尺度空间（CSS）角点检测中，通常采用高斯滤波对图像进行去噪处理。但该方法很大程度上降低了角点检测的准确性和精确性，因此，提出了利用双边滤波来改进的CSS角点检测算法，可以在很好去噪的同时又保证角点检测的高精度。实验结果也充分验证了该算法较其他算法具有的高效性和准确性。

关键词: 角点检测, 曲率尺度空间, 双边滤波, 高斯滤波