Robust Image Stitching with Optimal Plane

2025-08-07

AI Summary

To address the weak scene generalization and the inherent trade-off between content alignment and geometric structure preservation in unsupervised deep image stitching, this paper proposes RopStitch. Methodologically, it introduces a dual-branch network that jointly leverages semantic invariance and fine-grained features, integrating pretrained feature extraction with learnable correlation layers. It further pioneers the concept of a “virtual optimal plane,” modeling geometric consistency via homography decomposition and bidirectional planar mapping. Finally, an iterative coefficient predictor enables end-to-end optimization. Evaluated on multiple real-world datasets, RopStitch significantly improves stitching robustness and visual naturalness. Notably, it demonstrates superior generalization to unseen scenes compared to state-of-the-art methods, validating its effectiveness in challenging, unconstrained environments.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

We present extit{RopStitch}, an unsupervised deep image stitching framework with both robustness and naturalness. To ensure the robustness of extit{RopStitch}, we propose to incorporate the universal prior of content perception into the image stitching model by a dual-branch architecture. It separately captures coarse and fine features and integrates them to achieve highly generalizable performance across diverse unseen real-world scenes. Concretely, the dual-branch model consists of a pretrained branch to capture semantically invariant representations and a learnable branch to extract fine-grained discriminative features, which are then merged into a whole by a controllable factor at the correlation level. Besides, considering that content alignment and structural preservation are often contradictory to each other, we propose a concept of virtual optimal planes to relieve this conflict. To this end, we model this problem as a process of estimating homography decomposition coefficients, and design an iterative coefficient predictor and minimal semantic distortion constraint to identify the optimal plane. This scheme is finally incorporated into extit{RopStitch} by warping both views onto the optimal plane bidirectionally. Extensive experiments across various datasets demonstrate that extit{RopStitch} significantly outperforms existing methods, particularly in scene robustness and content naturalness. The code is available at {color{red}https://github.com/MmelodYy/RopStitch}.

Problem

• Robust unsupervised deep image stitching framework
• Dual-branch architecture for content perception integration
• Virtual optimal planes to resolve alignment-structure conflict

Innovation

• Unsupervised dual-branch deep learning architecture
• Virtual optimal planes with homography decomposition
• Bidirectional warping onto optimal planes