论文概要
Research Area: Computer Vision Authors: GeonU Kim, Shin Dong-Yeon, Tae-Hyun Oh Published: 2026-09-04 arXiv: 2609.05382
Abstract
We propose Ref-GeNVS, a training-free, reflection-aware method for generative novel view synthesis (NVS) in mirror scenes. Existing multi-view diffusion models often fail to recognize the mirror in the scene and cannot exploit reflected content for scene generation. To fix this issue without additional training, our key idea is to treat a mirror image as two complementary views.
From input images, we estimate the mirror plane and reflect camera poses to form virtual views. Based on this virtual view setup, we propose a two-stage generation method consisting of Mirror-gated attention and Reflection injection, which enables reflection-consistent NVS by explicitly leveraging reflection relationships in a multi-view diffusion model. Ref-GeNVS inherits the strong generalizability of the multi-view diffusion backbone without fine-tuning.
Key Contributions
- Training-free mirror handling: treats a mirror image as two complementary views instead of requiring retraining or fine-tuning.
- Virtual view construction: estimates the mirror plane from input images and reflects camera poses to form virtual views.
- Two-stage generation: Mirror-gated attention and Reflection injection explicitly exploit reflection relationships within the multi-view diffusion model for reflection-consistent synthesis.
- Strong generalization: inherits the generalization ability of the multi-view diffusion backbone with no additional training.
Results
On both synthetic and real scenes containing mirrors, Ref-GeNVS outperforms recent generative NVS methods by generating reflection-consistent and contextually coherent novel views, and can reveal scene structures that are only visible through the mirror.
--- *Auto-collected on 2026-09-09*