一种用于图像序列拼接的角点检测算法

分析了Harris角点检测算子的实现原理及其用于图像序列拼接时存在的不足,从几个方面提出了改进:用新的X和Y方向梯度算子求取图像导数,对灰度变化更敏感;改进了角点响应函数,避免了k值选择的随机性;根据图像序列首图像自动确定R阈值,提高了算法的自动处理能力和速度;根据目标像素点的8邻域范围进行亚像素精确定位,角点位置更加准确.实验结果表明,改进算法无需人工干预,特征点提取的速度比传统算法提高了约61.3%,能更好地适应图像拼接的实际应用,是一种快速有效的角点检测方法.

Corner Detection Algorithm for Image Mosaic

An improved algorithm for corner detection was proposed by analyzing the theory of Harris operator and its disadvantage. The main improved aspects were suggested just as follows ; First, the new grads operators were used to get image derivative. Second, the method adopted improved corner response function for avoiding the randomicity of k value. Then this algorithm automatically ascertained corner extraction threshold based on the first one in image sequence.Last, in order to improve the precision for corner extraction, the method obtained sub-pixel corners within 8 neighborhoods of the target pixel. Experiment results show that the computation time of improved algorithm enhances about 61. 3% , and it can be applied to image mosaic better.