SARA: Scene-Aware Reconstruction Accelerator

2026-01-11

arXiv.org

AI Summary

This work proposes a geometry-driven image pair selection mechanism to overcome the high computational cost and limited reconstruction accuracy of traditional Structure-from-Motion (SfM) methods, which rely on visual similarity for pairing. Prior to feature matching, the approach evaluates the informativeness of candidate pairs using overlap and disparity metrics, constructs an information-weighted spanning tree, and incorporates a key-edge enhancement strategy. This method introduces the first “geometry-first” pairing paradigm, substantially reducing matching complexity. Experiments demonstrate that, across multiple learning-based feature detectors, it achieves a 46.5% reduction in average rotation error, a 12.5% decrease in translation error, and a 98% reduction in the number of matched pairs, while accelerating the pipeline by up to 50×—all with reconstruction metric deviations maintained within ±3%.

Technology Category

• Computer Vision: Other Foundations of Computer Vision
• Machine Learning: Feature Construction/Reformulation
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

We present SARA (Scene-Aware Reconstruction Accelerator), a geometry-driven pair selection module for Structure-from-Motion (SfM). Unlike conventional pipelines that select pairs based on visual similarity alone, SARA introduces geometry-first pair selection by scoring reconstruction informativeness - the product of overlap and parallax - before expensive matching. A lightweight pre-matching stage uses mutual nearest neighbors and RANSAC to estimate these cues, then constructs an Information-Weighted Spanning Tree (IWST) augmented with targeted edges for loop closure, long-baseline anchors, and weak-view reinforcement. Compared to exhaustive matching, SARA reduces rotation errors by 46.5+-5.5% and translation errors by 12.5+-6.5% across modern learned detectors, while achieving at most 50x speedup through 98% pair reduction (from 30,848 to 580 pairs). This reduces matching complexity from quadratic to quasi-linear, maintaining within +-3% of baseline reconstruction metrics for 3D Gaussian Splatting and SVRaster.

Problem

• Structure-from-Motion
• pair selection
• reconstruction efficiency
• geometry-aware matching
• computational complexity

Innovation

• geometry-driven pair selection
• Information-Weighted Spanning Tree
• reconstruction informativeness
• Structure-from-Motion acceleration
• quasi-linear matching complexity