基于尺度聚类仿射过滤的图像拼接算法

Image Mosaic Algorithm Based on Scale-Clustering Affine Filtering

利用图像的SIFT特征点的位置和尺度组合成数据向量,通过对不同尺度特征点进行聚类分析得到具有较好稳定性的全局仿射变换.通过该变换对匹配特征点进行一致性过滤,并对图像拼接过程中出现的问题,构建了一种基于网格覆盖的拼接算法提高拼接效果和稳定性.通过实验对算法进行测试,在实验过程中,为度量过滤结果稳定性和边缘拼接效果,构建了相应的视觉和定量评价指标来对比算法性能.实验结果表明,提出的算法在图像拼接过程中具有较好的稳定性和拼接效果,具有较好的实用价值.

The positions of SIFT points and scale value were composed to a vector,which was used to the dataset of cluttering algorithm.The parameters of global affine transformation were got by the clustering algorithm,which was used to filter the matched point.For the image mosaic problems of the existing algorithms,a grid-selection mosaic algorithm was proposed.The visual skeleton and the image mosaic offset were put forward to measure the stability of the filtering and the performance of mosaic algorithm.Experimental results indicate that the methods proposed can overcome the problems of traditional algorithms and have better stability and accuracy.