Seamlessly Natural: Image Stitching with Natural Appearance Preservation

TL;DR

SENA tackles parallax-induced distortions in image stitching by replacing rigid homographies with a geometry-driven pipeline that combines a global affine initialization $A_{ ext{glob}}$, locally adaptive affine refinements, and a smoothly regularized Free-Form Deformation (FFD). It further identifies a parallax-minimized zone from disparity consistency without semantic preprocessing and enforces one-to-one geometry through anchor-based seam segmentation, supplemented by a dual-gated suppression mechanism to prevent ghosting and texture distortion. The method integrates a composite diagnostic-based selection of local transforms, confidence-weighted blending, and seam-aware reconstruction to yield panoramas with improved shape preservation and texture integrity. Experiments on challenging datasets show alignment comparable to leading homography-based methods while delivering significantly better visual realism and structural fidelity.

Abstract

This paper introduces SENA (SEamlessly NAtural), a geometry-driven image stitching approach that prioritizes structural fidelity in challenging real-world scenes characterized by parallax and depth variation. Conventional image stitching relies on homographic alignment, but this rigid planar assumption often fails in dual-camera setups with significant scene depth, leading to distortions such as visible warps and spherical bulging. SENA addresses these fundamental limitations through three key contributions. First, we propose a hierarchical affine-based warping strategy, combining global affine initialization with local affine refinement and smooth free-form deformation. This design preserves local shape, parallelism, and aspect ratios, thereby avoiding the hallucinated structural distortions commonly introduced by homography-based models. Second, we introduce a geometry-driven adequate zone detection mechanism that identifies parallax-minimized regions directly from the disparity consistency of RANSAC-filtered feature correspondences, without relying on semantic segmentation. Third, building upon this adequate zone, we perform anchor-based seamline cutting and segmentation, enforcing a one-to-one geometric correspondence across image pairs by construction, which effectively eliminates ghosting, duplication, and smearing artifacts in the final panorama. Extensive experiments conducted on challenging datasets demonstrate that SENA achieves alignment accuracy comparable to leading homography-based methods, while significantly outperforming them in critical visual metrics such as shape preservation, texture integrity, and overall visual realism.

Figures (23)

• Figure 1: Limitations of Existing Stitching Methods
• Figure 2: Flowchart of the proposed approach
• Figure 3: Optimal stitching line
• Figure 4: Complementary segments
• Figure 5: Complementary segments