摘要
针对目前图像匹配算法中存在的匹配精度不高和匹配速度慢的缺点,对基于灰度相关的2类匹配算法——最小误差法和相关系数法进行了改进。最小误差法采用新的ML距离法,提出动态调整阈值的方法,既保证了匹配精度,又避免了局部噪声的影响;相关系数法对相关系数的计算公式进行了简化,并采用三步搜索策略进行匹配,以达到减少计算量和搜索位置的目的。实验证明:改进后的算法,在保证一定匹配精度的条件下,匹配速度大大提高,能够满足实际应用中的实时性要求。
Gray correlation based minimum error and correlation coefficient algorithms were improved to eliminate the disadvantages of low matching accuracy and slow matching velocity existing in the image matching algorithms nowadays. In the improved minimum error algorithm, a new ML distance method is adopted and a threshold dynamic adjustment method is proposed to assure the matching accuracy and to avoid the influence of local noises. As for the correlation coefficient algorithm, the expression of correlation coefficient is simplified and three-step searching way is used to reduce t.he compexity of the calculation and to achieve the goal of the position acquisition. The experiment shows that the improved algorithms greatly increase the matching velocity without the loss of matching accuracy,and meet the requirement of real-time image matching system.
出处
《应用光学》
CAS
CSCD
2007年第5期536-540,共5页
Journal of Applied Optics
关键词
相关匹配
相关系数
匹配精度
匹配速度
correlation matching
correlation coefficient
matching accuracy
matching velocity
