Bracket Diffusion: HDR Image Generation by Consistent LDR Denoising

TL;DR

This paper introduces a novel method called Bracket Diffusion that generates HDR images by coordinating multiple pre-trained LDR diffusion models, leveraging a consistency term to fuse LDR brackets into a high dynamic range output.

Contribution

It presents a new approach to HDR image generation using multiple black-box LDR diffusion models with a consistency mechanism, avoiding the need for large HDR datasets or re-training.

Findings

Achieves state-of-the-art HDR generation results

Effectively fuses multiple LDR brackets into HDR images

Demonstrates both unconditional and conditional HDR synthesis

Abstract

We demonstrate generating HDR images using the concerted action of multiple black-box, pre-trained LDR image diffusion models. Relying on a pre-trained LDR generative diffusion models is vital as, first, there is no sufficiently large HDR image dataset available to re-train them, and, second, even if it was, re-training such models is impossible for most compute budgets. Instead, we seek inspiration from the HDR image capture literature that traditionally fuses sets of LDR images, called "exposure brackets'', to produce a single HDR image. We operate multiple denoising processes to generate multiple LDR brackets that together form a valid HDR result. The key to making this work is to introduce a consistency term into the diffusion process to couple the brackets such that they agree across the exposure range they share while accounting for possible differences due to the quantization error. We demonstrate state-of-the-art unconditional and conditional or restoration-type (LDR2HDR) generative modeling results, yet in HDR.