Scene-Consistent Illumination Transfer for Inserted Advertising Graphics

2026-09-29

AI Summary

This study addresses the visual incongruity in advertising graphic insertion caused by inconsistent lighting and shadows, proposing Ad-Relight. This method introduces a novel training-free, inference-only pipeline that leverages pretrained diffusion-based relighting models for scene illumination transfer. By employing dual-background probing to isolate the target region's lighting contributions, combined with structure disentanglement, residual fusion, soft-decay masking, and smooth luminance field optimization, it achieves high-fidelity image synthesis. Experiments on 560 cases demonstrate that the proposed approach significantly improves SSIM and perceptual consistency, outperforming existing baselines particularly under non-uniform illumination conditions, thereby enabling photorealistic advertising compositing.

Abstract

Replacing a visible advertisement in a broadcast frame is geometrically straightforward but photometrically delicate. A pasted graphic can have the correct perspective and still appear detached when its brightness, shading, or shadow disagrees with the surface beneath it. This paper presents Ad-Relight, an inference-only procedure for transferring scene illumination to a supplied advertising graphic without collecting a banner-specific training set. The procedure first separates slowly varying shade from graphic structure, then probes a pretrained diffusion relighter with two nearly identical backgrounds to isolate the contribution of the target region. A final pass combines this residual with a smoothed luminance field and a soft attenuation mask. Across 560 generated placements, the approach improves structural similarity, perceptual distance, and illumination agreement over geometric compositing and direct relighting baselines. Human judgments and an automated preference study show the clearest gains on floor-mounted graphics with nonuniform lighting. The current study is image based; temporal stabilization remains an open extension.

Problem

• illumination transfer
• advertising insertion
• scene-consistent relighting
• image compositing
• photometric consistency

Innovation

• Illumination Transfer
• Diffusion Relighter
• Inference-only
• Scene-Consistent Compositing
• Ad-Relight