基于SURF特征相关的红外图像自动拼接

Infrared Image Automatic Mosaic Method Based on SURF Feather Correlation

提出一种鲁棒性较好的图像拼接算法解决红外图像序列的自动拼接问题。首先应用图像子区域相位相关得到图像间近似重合子区域和图像间近似仿射变换参数;接着根据得到的参数值对SURF特征计算和特征匹配进行优化,提高了图像配准速度;然后以特征点对的对称投影位置误差最小化为目标,应用RANSAC和LM算法对图像间透视变换矩阵进行优化;最后应用得到的透视变换建立图像序列整体配准关系,并采用线性过渡法进行重叠区域融合得到最终的拼接图像。

A robust image stitching algorithm is presented to solve the infrared image mosaic problem. Firstly, the sub-image phase correlation method was applied for similar sub-area searching and approximate affine transform calculating. Then, the SURF feather calculating and feather matching were optimized by the previous calculated affine parameters, and the speed of image alignment was increased. The symmetry projection error of SURF point pairs was mini- mized to optimize the projection matrix by RANSAC and LM method. Finally, the whole align- ment relationship of image sequence was obtained by these matrixes and the overlapped area be- tween images was fused by linear transition method to achieve the final mosaic image.