Interaction-Consistent Object Removal via MLLM-Based Reasoning

TL;DR

This paper introduces a new framework called REORM that uses multimodal large language models to perform interaction-consistent object removal in images, addressing the challenge of removing related interaction elements for more semantically coherent results.

Contribution

The paper formalizes the Interaction-Consistent Object Removal problem and proposes REORM, a novel reasoning-enhanced framework leveraging MLLMs for joint removal of objects and interaction elements.

Findings

REORM outperforms existing image editing systems on the ICOREval benchmark.

REORM effectively removes interaction elements, ensuring semantic consistency.

The modular design of REORM enables accurate editing even with limited resources.

Abstract

Image-based object removal often erases only the named target, leaving behind interaction evidence that renders the result semantically inconsistent. We formalize this problem as Interaction-Consistent Object Removal (ICOR), which requires removing not only the target object but also associated interaction elements, such as lighting-dependent effects, physically connected objects, targetproduced elements, and contextually linked objects. To address this task, we propose Reasoning-Enhanced Object Removal with MLLM (REORM), a reasoningenhanced object removal framework that leverages multimodal large language models to infer which elements must be jointly removed. REORM features a modular design that integrates MLLM-driven analysis, mask-guided removal, and a self-correction mechanism, along with a local-deployment variant that supports accurate editing under limited resources. To support evaluation, we introduce ICOREval, a benchmark consisting of instruction-driven removals with rich interaction dependencies. On ICOREval, REORM outperforms state-of-the-art image editing systems, demonstrating its effectiveness in producing interactionconsistent results.