一种快速鲁棒性的视差图像拼接方法

A Fast and Robust Stitching Method for Parallax Images

当小型旋翼无人机超低空飞行时,不同位置拍摄的图像中极易出现结构视差。传统基于全局投影的拼接方法在处理视差图像时会出现伪影或畸变。最近的自适应拼接方法,如As-Projective-As-Possible(APAP)及其优化方法,使用“网格”来构建密集的局部投影场域,以改善视差图像拼接的对齐度和自然度,但计算复杂度较大。为此,提出了一种新颖的视差图像拼接方法,以深度稠密投影(Deep Dense Projection,Deep-DP)为特征匹配框架,设计了一种轻量化并行计算框架求解平滑形变场,以实现视差图像无缝拼接。在对比试验中,该算法平均运行时间达到0.2 s,显著优于其他先进算法,并且对齐质量和自然度达到了相近水平。提出的算法在兼顾图像拼接准确性的同时,计算效率显著提高。

When small rotary-wing unmanned aerial vehicles fly at ultra-low altitudes,images captured at different positions often suffer from significant parallax.Traditional global projection-based stitching methods tend to produce artifacts or distortions when dealing with parallax images.Recent adaptive stitching methods,such as As-Projective-As-Possible(APAP)and its optimization techniques,utilize“grid”to construct dense local projection fields to improve the alignment and naturalness of parallax image stitching,but they are computationally complex.In order to achieve this,a novel method for stitching parallax images is proposed,utilizing Deep Dense Projection(Deep-DP)as the feature matching framework.A lightweight parallel computing framework is designed to solve the smooth warping field,enabling seamless stitching.In comparative experiments,the proposed algorithm achieves an average runtime of 0.2 s,significantly outperforming other state-of-the-art algorithms,while attaining alignment accuracy and naturalness at a similar level.These results indicate that the proposed algorithm significantly enhances computational efficiency while ensuring image stitching accuracy.