Lookup Table meets Local Laplacian Filter: Pyramid Reconstruction Network for Tone Mapping

TL;DR

This work tackles HDR to LDR tone mapping by addressing the limitations of global 3D LUTs in capturing local details. It introduces a joint global-local framework that leverages a reversible Laplacian pyramid to separate low-frequency tonal adjustments (via image-adaptive 3D LUTs) from high-frequency edge preservation (via a learnable local Laplacian filter guided by a lightweight transformer). The method optimizes a composite loss that includes reconstruction, perceptual, and regularization terms, with per-pixel LUT fusion and adaptive LLF parameter maps enabling end-to-end training. Experiments on HDR+ and MIT-FiveK across 480p and 4K demonstrate clear performance gains over state-of-the-art approaches, while maintaining computational efficiency suitable for high-resolution imagery.

Abstract

Tone mapping aims to convert high dynamic range (HDR) images to low dynamic range (LDR) representations, a critical task in the camera imaging pipeline. In recent years, 3-Dimensional LookUp Table (3D LUT) based methods have gained attention due to their ability to strike a favorable balance between enhancement performance and computational efficiency. However, these methods often fail to deliver satisfactory results in local areas since the look-up table is a global operator for tone mapping, which works based on pixel values and fails to incorporate crucial local information. To this end, this paper aims to address this issue by exploring a novel strategy that integrates global and local operators by utilizing closed-form Laplacian pyramid decomposition and reconstruction. Specifically, we employ image-adaptive 3D LUTs to manipulate the tone in the low-frequency image by leveraging the specific characteristics of the frequency information. Furthermore, we utilize local Laplacian filters to refine the edge details in the high-frequency components in an adaptive manner. Local Laplacian filters are widely used to preserve edge details in photographs, but their conventional usage involves manual tuning and fixed implementation within camera imaging pipelines or photo editing tools. We propose to learn parameter value maps progressively for local Laplacian filters from annotated data using a lightweight network. Our model achieves simultaneous global tone manipulation and local edge detail preservation in an end-to-end manner. Extensive experimental results on two benchmark datasets demonstrate that the proposed method performs favorably against state-of-the-art methods.

Figures (6)

• Figure 1: Overview of the framework. Our method first decomposes the input image $\mathbf{I}$ into a Laplacian pyramid. The low-frequency image $\mathbf{I}_{low}$ is fed into a lightweight transformer weight predictor and Basis 3D LUTs fusion block to transform into a low-resolution enhanced image $\hat{\mathbf{I}}_{low}$. To adaptively refine the high-frequency components, we progressively learn an image-adaptive local Laplacian filter (LLF) based on both high- and low-frequency images. Then, we perform the remapping function of the local Laplcian filter to refine the high-frequency components while preserving the pyramid reconstruction capability. For the level $N-1$, we concatenate the component with the edge map of $\hat{\mathbf{I}}_{low}$ to mitigate potential halo artifacts.
• Figure 2: Illustration of the basis 3D LUTs fusion strategy. (a) present the multiple pixel mapping relationships of an image pair; (b) is the conventional basis 3D LUTs fusion strategy; (c) is the pixel-level basis 3D LUTs fusion strategy.
• Figure 3: Architecture of the proposed image-adaptive learnable local Laplacian filter (LLF).
• Figure 4: Visual comparison with state-of-the-art methods on a test image from the HDR+ dataset hasinoff2016burst. The error maps in the upper left corner facilitates a more precise determination of performance differences. Best viewed in color and by zooming in.
• Figure 5: Visual comparison with state-of-the-art methods on a test image from the MIT-Adobe FiveK dataset bychkovsky2011learning. The error maps in the upper left corner facilitates a more precise determination of performance differences. Best viewed in color and by zooming in.