【摘要】 针对无先验信息传统算法中普遍存在的误差累计问题,提出基于加窗尺度不变特征变换(W-SIFT)和分布式优化的多图自动拼接算法。根据多图拼接应用的特性,对尺度不变特征变换算法进行修改,提出加窗SIFT算法更高效地提取待拼接图像的特征点。运用随机抽样一致(RANSAC)算法计算出两两图像的变换矩阵。之后,建立了一个分布式优化模型,求解出多图拼接的全局最优解。实验结果表明,基于加窗SIFT和分布式优化的多图自动拼接算法能够有效地消除误差累积现象,能够得到更加精确的多图拼接结果。更多还原

【Abstract】 Aiming at the error accumulation problem existing in the traditional algorithms without prior information,a multiimage automatic splicing algorithm based on windowed scale invariant feature transformation(W-SIFT)and distributed optimization is proposed. The scale invariant feature transformation algorithm is modified according to the application characteristics of the multi-image splicing. The W-SIFT algorithm is proposed to extract the feature points of the splicing image efficiently. The random sample consensus(RANSAC)method is used to calculate the transformation matrix of two images. A distributed optimization model was established to solve the global optimal solution of the multi-image splicing. The experimental results show that the multi-image automatic splicing algorithm based on W-SIFT and distributed optimization can eliminate the error accumulation phenomenon effectively,and obtain the accurate multi-image splicing results.更多还原