GeoRemover: Removing Objects and Their Causal Visual Artifacts

2025-09-22

AI Summary

Existing image editing methods struggle to eliminate causal visual artifacts—such as shadows and reflections—during object removal: strict mask alignment leaves residual artifacts, while loose alignment causes over-erasure and reduced controllability. This stems from neglecting the causal relationship between an object’s geometric presence and its visual effects. To address this, we propose a geometry-aware two-stage editing framework. In the first stage, object geometry is explicitly removed via depth estimation; in the second stage, appearance is conditionally generated. Our approach is the first to decouple and model the geometry–vision causal dependency, integrating mask-alignment supervision with preference-driven learning (trained on positive/negative sample pairs) to jointly ensure structural fidelity and editing controllability. Evaluated on two mainstream benchmarks, our method achieves state-of-the-art performance, significantly improving artifact removal quality while avoiding structural distortion and spurious deletion of irrelevant regions.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Machine Learning: Causal Learning

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Algorithmic accountability and transparency on the web

Abstract

Towards intelligent image editing, object removal should eliminate both the target object and its causal visual artifacts, such as shadows and reflections. However, existing image appearance-based methods either follow strictly mask-aligned training and fail to remove these causal effects which are not explicitly masked, or adopt loosely mask-aligned strategies that lack controllability and may unintentionally over-erase other objects. We identify that these limitations stem from ignoring the causal relationship between an object's geometry presence and its visual effects. To address this limitation, we propose a geometry-aware two-stage framework that decouples object removal into (1) geometry removal and (2) appearance rendering. In the first stage, we remove the object directly from the geometry (e.g., depth) using strictly mask-aligned supervision, enabling structure-aware editing with strong geometric constraints. In the second stage, we render a photorealistic RGB image conditioned on the updated geometry, where causal visual effects are considered implicitly as a result of the modified 3D geometry. To guide learning in the geometry removal stage, we introduce a preference-driven objective based on positive and negative sample pairs, encouraging the model to remove objects as well as their causal visual artifacts while avoiding new structural insertions. Extensive experiments demonstrate that our method achieves state-of-the-art performance in removing both objects and their associated artifacts on two popular benchmarks. The code is available at https://github.com/buxiangzhiren/GeoRemover.

Problem

• Removing objects and their causal visual artifacts in images
• Addressing limitations of existing appearance-based object removal methods
• Proposing a geometry-aware two-stage framework for structure-aware editing

Innovation

• Geometry-aware two-stage framework decouples object removal
• Strictly mask-aligned geometry removal with structural constraints
• Geometry-conditioned appearance rendering implicitly handles visual artifacts