Learning Residual Elastic Warps for Image Stitching under Dirichlet Boundary Condition

TL;DR

This paper introduces REwarp, a novel recurrent elastic warp method that predicts homography and TPS under Dirichlet boundary conditions, effectively reducing holes and discontinuities in image stitching, and demonstrating competitive performance.

Contribution

The paper proposes REwarp, a new elastic warping approach with boundary constraints and residual learning for improved image stitching quality.

Findings

REwarp effectively reduces holes and discontinuities.

It achieves competitive alignment accuracy.

The method maintains reasonable computational costs.

Abstract

Trendy suggestions for learning-based elastic warps enable the deep image stitchings to align images exposed to large parallax errors. Despite the remarkable alignments, the methods struggle with occasional holes or discontinuity between overlapping and non-overlapping regions of a target image as the applied training strategy mostly focuses on overlap region alignment. As a result, they require additional modules such as seam finder and image inpainting for hiding discontinuity and filling holes, respectively. In this work, we suggest Recurrent Elastic Warps (REwarp) that address the problem with Dirichlet boundary condition and boost performances by residual learning for recurrent misalign correction. Specifically, REwarp predicts a homography and a Thin-plate Spline (TPS) under the boundary constraint for discontinuity and hole-free image stitching. Our experiments show the favorable aligns and the competitive computational costs of REwarp compared to the existing stitching methods. Our source code is available at https://github.com/minshu-kim/REwarp.