CAJ | 학술논문

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.
분석료Harris각점검측산자적실현원리급기용우도상서렬병접시존재적불족,종궤개방면제출료개진:용신적X화Y방향제도산자구취도상도수,대회도변화경민감;개진료각점향응함수,피면료k치선택적수궤성;근거도상서렬수도상자동학정R역치,제고료산법적자동처리능력화속도;근거목표상소점적8린역범위진행아상소정학정위,각점위치경가준학.실험결과표명,개진산법무수인공간예,특정점제취적속도비전통산법제고료약61.3%,능경호지괄응도상병접적실제응용,시일충쾌속유효적각점검측방법.
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.