摘要: 在序列图像匹配应用中,最小亮度变化(MIC)算法存在角点漏检、对噪声敏感、不具有旋转不变性等缺点,降低了特征点检测的精确性和快速性。针对该问题,对MIC算法进行改进,并与加速稳健特征(SURF)算法相结合,提出一种快速的图像匹配算法。对图像进行自适应平滑滤
波,在图像的非平坦区域运用8邻域像素相似法检测需要的几何角点,并采用SURF算法对检测到的特征点进行描述和匹配。实验结果表明,与SURF算法、最小核值相似区-加速稳健特征算法相比,该算法在图像发生旋转变化、噪声变化、亮度变化和仿射变化时都可以提高图像
匹配的速度和准确率。
关键词:
特征点检测,
图像匹配,
最小亮度变化,
加速稳健特征算法,
8邻域像素
Abstract: In sequence image matching application,the disadvantages of the Minimum Intensity Change(MIC) algorithm,such as having no rotational invariance,missing many corners and being susceptible to noise,reduce the accuracy and spead of feature point detection.In order to solve this problem,the MIC algorithm is improved,and a fast image matching algorithm is proposed in conjunction with Speed Up Robust Feature(SURF) algorithm.Self-adaption smoothing filtering is processed in images,then by applying 8-neighborhood pixel similarity method,necessary geometric corners are detected in the non-flat areas of the image,finally after SURF algorithm is applied,feature points detected are described and matched.Experimental results indicate that compared with SURF and Smallest Univalue Segment Assimilating Nucleus-speed Up Robust Feature(SUSAN-SURF),this algorithm can improve the speed and accuracy of image matching under the changes about image rotation,noise,intensity and affinity.
Key words:
feature point detection,
image matching,
Minimum Intensity Change(MIC),
Speed up Robust Feature(SURF) algorithm,
8-neighborhood pixel
中图分类号: