Abstract:Irregular boundaries in image stitching naturally occur due to freely moving cameras. To deal with this problem, existing methods focus on optimizing mesh warping to make boundaries regular using the traditional explicit solution. However, previous methods always depend on hand-crafted features (e.g., keypoints and line segments). Thus, failures often happen in overlapping regions without distinctive features. In this paper, we address this problem by proposing RecStitchNet, a reasonable and effective network for image stitching with rectangular boundaries. Considering that both stitching and imposing rectangularity are non-trivial tasks in the learning-based framework, we propose a three-step progressive learning based strategy, which not only simplifies this task, but gradually achieves a good balance between stitching and imposing rectangularity. In the first step, we perform initial stitching by a pre-trained state-of-the-art image stitching model, to produce initially warped stitching results without considering the boundary constraint. Then, we use a regression network with a comprehensive objective regarding mesh, perception, and shape to further encourage the stitched meshes to have rectangular boundaries with high content fidelity. Finally, we propose an unsupervised instance-wise optimization strategy to refine the stitched meshes iteratively, which can effectively improve the stitching results in terms of feature alignment, as well as boundary and structure preservation. Due to the lack of stitching datasets and the difficulty of label generation, we propose to generate a stitching dataset with rectangular stitched images as pseudo-ground-truth labels, and the performance upper bound induced from the it can be broken by our unsupervised refinement. Qualitative and quantitative results and evaluations demonstrate the advantages of our method over the state-of-the-art.

DocUNet: Document Image Unwarping Via a Stacked U-Net

DewarpNet: Single-Image Document Unwarping with Stacked 3D and 2D Regression Networks.

UVDoc: Neural Grid-based Document Unwarping

Learning from Documents in the Wild to Improve Document Unwarping

Rethinking Supervision in Document Unwarping: A Self-consistent Flow-free Approach

Effective Document Image Rectification via a Deep Learning Framework

RecStitchNet: Learning to Stitch Images with Rectangular Boundaries

Layout-aware Single-image Document Flattening

Document Registration: Towards Automated Labeling of Pixel-Level Alignment Between Warped-Flat Documents

Deep Unrestricted Document Image Rectification.

DocScanner: Robust Document Image Rectification with Progressive Learning

Foreground and Text-lines Aware Document Image Rectification

Restoring Camera-Captured Distorted Document Images

Efficient Joint Rectification of Photometric and Geometric Distortions in Document Images.

Fourier Document Restoration for Robust Document Dewarping and Recognition

Transformer-Based UNet with Multi-Headed Cross-Attention Skip Connections to Eliminate Artifacts in Scanned Documents

DocTr: Document Image Transformer for Geometric Unwarping and Illumination Correction

Adaptive dewarping of severely warped camera-captured document images based on document map generation

Geometric Representation Learning for Document Image Rectification

MataDoc: Margin and Text Aware Document Dewarping for Arbitrary Boundary

UDRSNet: An unsupervised deformable registration module based on image structure similarity