摘要
为提高红外图像拼接速度和精度,对基于特征点匹配的图像拼接算法进行改进。根据图像空间特性减小角点搜索范围,通过设定梯度阈值,对梯度超过阈值的像素点进行Harris角点检测;改进Harris角点响应函数和角点筛选阈值的设定方式,摆脱了角点检测对筛选经验值的依赖。在相似测度Normalized Cross Correlation(NCC)粗匹配的基础上,采用有约束条件的随机选取方式,增强子集选取的合理性;并根据先局部后整体的匹配策略,基于匹配点的特性进行预检验,降低匹配错误率。算法最后利用最优变换矩阵确定待拼接图像的位置关系,实现自动拼接。实验结果表明,改进后算法在拼接过程中无需人工干预,在保证红外图像拼接质量的基础上,拼接速度提高了65.92%。
In order to improve the speed and the accuracy of infrared image mosaicing, we made an improvement on the feature point matching-based image mosaicing algorithm. The scanning range of the comer detection was narrowed according to the spatial feature of the iIoage, and by setting the gradient threshold, Harris comer detection was applied to those pixel points with gradient exceeding the threshold; The Harris comer responding function and the setting way of comer screening threshold were improved, this get rid of the dependence of comer detection on screening experience value. Based on similarity metric NCC (normalised cross correlation) coarse matching, we adopted the constraint random selection means to enhance the rationality of subset selection ; and according to local-to-global matching strategy we made pre-detection based on the features of matching points to lower the matching error rate. At last the algorithm used the optimal transformation matrix to determine the position relationships between images under mosaicing, thus realised auto-mosaicing. Experimental results showed that the improved algorithm did not need artificial intervention in mosaicing process, the mosaicing speed increased by 65.92% on the basis of ensuring the quality of infrared images mosaicing.
出处
《计算机应用与软件》
CSCD
2015年第9期192-196,共5页
Computer Applications and Software
基金
国家自然科学基金项目(61171126)
上海市自然科学基金项目(11ZR1415200)
上海重点支撑项目(12250501500)