基于双边滤波的Harris角点检测

Harris Corner Detection Based on Bilateral Filtering

角点是图像重要的特征之一。角点检测在图像处理和计算机视觉中起着重要的作用。原始的Harris角点检测不具有尺度不变性且常常将噪声作为角点,为了解决此问题,结合尺度空间理论和双边滤波的思想,对原始的Harris算法进行了改进。改进的算法加入高斯核,让算法具有尺度性,加入双边滤波,对检测的对象去噪保边,提高了检测精度。利用该算法,也可以快速准确地自动实现了对一定区域的棋盘格角点检测与提取。实验证明,该改进算法有较高的检测精度、准确率和稳定性,达到了预期的效果。

The corner is one of the important features of images. Corner-detection is a common operator which is used in image processing and computer vision. The original Harris corner detection does not have the property of scale invariant and often takes the noise for corners. In order to solve the above problems, the idea of combination of scale-space and bilateral filter is put forward and applied in the original Harris algorithm so that the algorithm has been improved. The improved algorithm has the property of scale invariance when Gaussian kernel added in. With the application of bilateral filtering, the noise can be removed from the object and the integrity of edge can be protected well. Importantly, the detection accuracy has been improved greatly. Checkerboard corners detection and extraction also can be achieved automatically with the improved algorithm. The experiment shows that this method is more strict and more accurate for the detection objects and reduces the false rate of corner detection greatly, which achieves the desired result.