一种鲁棒的序列图像自动拼接方法

针对传统图像拼接方法中鲁棒性差、计算量大及自动化程度低等问题，提出一种鲁棒性高的序列图像自动拼接方法。该方法首先采用Harris角点检测算子对经Wallis滤波后的序列图像进行特征点提取，并结合Forstner算子对特征点进行精确定位。然后基于所提取的特征点,采用邻域灰度互相关法进行序列图像的特征点匹配，得到粗匹配点集，并运用RANSAC算法对粗匹配点集处理得到精匹配点集，由精匹配点集求出较高精度的基础矩阵及极线，并由极线约束引导匹配得到高精度的匹配点对，再运用双向松弛整体匹配算法进一步剔除少数位于极线上的误匹配点。最后利用所得的高精度匹配点对，求解序列图像间的仿射变换关系，并进行图像的坐标变换和融合，从而实现序列图像的自动拼接。实验结果表明，该方法拼接效果理想，鲁棒性高，整个拼接过程全自动，不需要人工干预，具有较高的实用价值。

Robust image sequence auto-mosaic method

Considering the low robustness,large computation and low automation in the traditional image mosaic method,a robust automatic image sequence mosaic method was proposed.First,the method used Harris operator and Forstner operator to extract the image character points after using Wallis filter to enhance the images.Then the gray cross correlation of the certain domain of the character points was utilized to obtain the rough matching points,and the RANSAC algorithm was used to get accurate image matching points.After that,the matching points were used to calculate the foundation matrix and epipolar lines,and the epipolar constraint was used to induct the image matching and get the final high precise matching points.Hence,bidirectional relaxation whole matching algorithm was used to eliminate the error matching points located on the epipolar line.Finally,the accurate matching points were utilized to calculate the affine transform relations of the image sequence and thus completed the image sequence auto-mosaic.The experimental results show that the proposed method is effective in image sequence mosaic,and the process is automatic with high practical value.