Reusing previous views speeds up interactive scene exploration
LVS: Local View Synthesis from Relative Camera Pose by Reusing Previous Views
Computer Vision and Pattern RecognitionArtificial IntelligenceGraphicsMultimedia
Summary
When exploring a 3D scene on a computer, moving the camera a little usually shows mostly the same things. The authors found a way to use images from previous views with some smart adjustments instead of redrawing every new view from scratch. They combine simple geometric warping with a small neural network to fix errors and add missing details. This makes the scene update faster while still looking good, helping in applications like virtual and augmented reality.
What this means in practice
- •For augmented reality developers: Create faster view updates by reusing previous images adjusted by camera motion, improving user experience without full scene rerendering.
- •For virtual reality content creators: Reduce latency during user navigation by using relative camera pose to synthesize nearby views quickly with minimal rendering.
Authors
Qizhou Huo, Xuan Sun, Yongfei Guo, Zhipeng Wang, Yuanhao Gong
Abstract
Interactive scene exploration requires frequent view updates, although small camera motions preserve much of the visible content. Conventional 3D Gaussian Splatting nevertheless renders each target view, leaving this image overlap unexploited. Reusing rendered images offers an alternative. Geometric warping alone cannot recover newly exposed content and remains sensitive to depth errors. We propose a per-scene framework that replaces repeated scene rendering for nearby views with relative-pose-guided RGB-D image reuse. Geometric warping uses depth and relative pose to transport source content, while a lightweight multiscale network predicts RGB residuals to correct artifacts and infer missing appearance. Cached source features further reduce repeated computation. On GS-render, residual refinement improves PSNR by 0.72~dB over pure warping; evaluations on captured and rendered scenes demonstrate low query latency. This separation of scene rendering from local view updates supports responsive scene exploration, with potential applications in augmented and virtual reality.