阻尼改进抗差L-M匹配下全景图像拼接平滑优化

Smooting optimization of panoramic image stitching with improved damping and anti-error L-M matching

为了解决大面积视觉全景图像受拼接缝影响,拼接平滑面积较小的问题,提出阻尼改进抗差L-M匹配下全景图像拼接平滑优化方法。阻尼因子改进抗差L-M算法,计算抗差L-M迭代向量;利用改进抗差L-M算法,融合球面投影理论,对特征匹配误差目标函数进行优化。融合拉普拉斯算法和高斯滤波金字塔对图像进行分解重构,确定图像拼接缝重叠区域后完成全景图像拼接平滑优化。实验结果表明,所提方法的拼接缝平滑面积能够达到95.8%,解决了拼接平滑面积较小的问题。

In order to solve the problem of large area visual panoramic images being affected by stitching and having a small stitching smooth area,a damping improved anti-error L-M matching based panoramic image stitching smooth optimization method is proposed.Improving the anti-error L-M algorithm with damping factor and calculating the anti-error L-M iteration vector;By utilizing the improved anti-error L-M algorithm and integrating the theory of spherical projection,the objective function of feature matching error is optimized.The fusion of Laplace algorithm and Gaussian filtering pyramid is used to decompose and reconstruct the image,determine the overlapping area of the image stitching seam,and complete the panoramic image stitching smoothing optimization.The experimental results show that the proposed method can achieve a smooth stitching area of 95.8%,solving the problem of small stitching areas.