Geometric Reciprocity: Unlocking Self-Supervision for Stereoscopic Video Generation
2026-07-06 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors developed a new method to turn regular 2D videos into 3D by creating stereo images without needing lots of paired 3D data. They introduced a rule called the Geometric Reciprocity Theorem that helps find which parts of the image are missing when shifting views, using only a single camera's video. This allows their system to teach itself from regular videos, improving the quality of 3D image creation without relying on expensive or limited training data. Their approach works better than other methods that either need no training or require supervised data.
Monocular-to-stereo conversionStereoscopic contentDepth-Image-Based Rendering (DIBR)DisocclusionStereo inpaintingSelf-supervised learningCycle consistencyGeometric Reciprocity TheoremNearest-neighbor warpingMonocular videos
Authors
Jingyi Lu, Kai Han
Abstract
Monocular-to-stereo conversion synthesizes stereoscopic content from 2D videos for immersive 3D experiences. In modern Depth-Image-Based Rendering (DIBR) approaches, stereo inpainting of disocclusions is the critical bottleneck. Training-based methods achieve superior quality but rely on scarce stereo pairs or synthetic data with domain gaps. We address this through the first self-supervised framework learning from monocular videos via cycle consistency. Our key contribution is the Geometric Reciprocity Theorem (GRT): under the nearest-neighbor DIBR formulation, the disocclusion mask when synthesizing a target view equals the mask of pixels lost when warping back from target to source, enabling analytical computation of test-time disocclusion masks directly from monocular images. This yields train-test consistency for the stated warping formulation, supporting self-supervised learning from unlimited monocular videos and substantial improvements over training-free and supervised state-of-the-art methods. Project page: https://visual-ai.github.io/grt/