摘要
传统角点检测算法对尺度很敏感,而且提取角点是像素级的。采用图像增强技术,通过DOG算子将多尺度运用到Harris算法中,然后除去极值点附近低对比度的特征点。不仅避免了传统灰度变换技术的单一性,还提高了增强处理后图像的稳定性。改进的多尺度Harris角点检测方法具有误差较小、伪角点较少、错误率较低、匹配精度性较高等特点。
Traditional corner detection algorithm is very sensitive to scale, and extract the corner point is the pixel-level. Using image enhancement technology, Through the DOG operator will be applied to the multi-scale Harris algorithm, Then remove the extreme point of the feature points near the low-contrast, To avoid the traditional gray-scale transformation technology singularity, Improved to enhance the stability of processed images. Experimental results show that the improved multi-scale Harris corner detection method has the charateristic of smaller error, pseudo-corner point less, error rate lower and the markedly improved accuracy and soon.
出处
《电脑开发与应用》
2010年第6期31-33,共3页
Computer Development & Applications
关键词
角点检测
多尺度
DOG算子
高斯金字塔
图像增强
corner detection, multi-scale, DOG operator, DOG pyramid, image enhancement