Multi-feature image stitching method with edge curve matching

Accurate image alignment remains a challenging problem in current panoramic stitching research. Most existing methods estimate warping models based on point and line feature correspondences, ignoring edge curve structures that effectively reflect the critical structure of the scene. To address this limitation, this paper proposes a multi-feature image stitching method that integrates an edge curve matching strategy to enhance the robustness of alignment. Initially, we derive a curve–point projective invariant from the characteristic number and develop an irregular edge curve matching strategy to obtain reliable curve correspondences. New intersection point feature pairs are then constructed from these correspondences to provide additional alignment constraints. Furthermore, a multi-feature guided warping model jointly constrained by point, line, and edge curve is employed to achieve precise alignment. Extensive experimental results demonstrate that the proposed method produces more accurate alignment, achieving improvements of 0.75% in SSIM and 3.2% in PSNR compared with state-of-the-art methods.