INRetouch: Context Aware Implicit Neural Representation for Photography Retouching

TL;DR

InRetouch presents a one-shot, context-aware implicit neural representation for photography retouching that learns from a single before-after pair and applies complex, region-specific edits to new images. The method employs window-based processing, a split RGB/position input design, and a lightweight context module to capture local structure while maintaining efficiency, enabling real-time 4K editing. A new Neural Retouch Dataset with 100k edited RAW images and 170 Lightroom presets supports robust evaluation and benchmarking, with demonstrated improvements in retouch transfer, Gamut Mapping, and RAW reconstruction. Ablations confirm the critical roles of context awareness, window sampling, and direct RGB input, while the approach remains limited by reference similarity and noise handling, suggesting directions for further refinement and broad adoption.

Abstract

Professional photo editing remains challenging, requiring extensive knowledge of imaging pipelines and significant expertise. While recent deep learning approaches, particularly style transfer methods, have attempted to automate this process, they often struggle with output fidelity, editing control, and complex retouching capabilities. We propose a novel retouch transfer approach that learns from professional edits through before-after image pairs, enabling precise replication of complex editing operations. We develop a context-aware Implicit Neural Representation that learns to apply edits adaptively based on image content and context, and is capable of learning from a single example. Our method extracts implicit transformations from reference edits and adaptively applies them to new images. To facilitate this research direction, we introduce a comprehensive Photo Retouching Dataset comprising 100,000 high-quality images edited using over 170 professional Adobe Lightroom presets. Through extensive evaluation, we demonstrate that our approach not only surpasses existing methods in photo retouching but also enhances performance in related image reconstruction tasks like Gamut Mapping and Raw Reconstruction. By bridging the gap between professional editing capabilities and automated solutions, our work presents a significant step toward making sophisticated photo editing more accessible while maintaining high-fidelity results. The source code and the dataset are publicly available at https://omaralezaby.github.io/inretouch .

Figures (15)

• Figure 1: We propose InRetouch, a novel implicit neural representation method for one-shot image retouching transfer. Our method learns the style from a single before-after pair, and transfers it to any given image. Unlike previous methods, InRetouch is not limited to global color editing, and can transfer a wide variety of edits, including region/object-specific edits. Additionally, our efficient method allows real-time 4K processing without artifacts. All outputs shown (2,3 raw) are produced by our method.
• Figure 2: Samples from our Neural Retouch Dataset. We can appreciate the diversity of styles and the challenging transformations.
• Figure 3: Our proposed InRetouch pipeline (A). Our method allows to learn complex photography edits from a single pair of before-after images. Our window sampling allows for fast optimization without constraints on the image size. Moreover, it enables including information from neighboring pixels while maintaining the simplicity and efficiency of traditional INRs. During inference, our model generalizes and transfers the edits to any input image. The full diagram of our proposed INR Architecture shown in (B).
• Figure 4: Comparison between different methods on retouching transfer task. Our method learns the edits effectively from a single sample, generalizing to a wide variety of edits, and has the most consistent output with the GT. We can appreciate the ability of our method to learn and adapt to complex edits like vignetting and local modification.
• Figure 5: Importance of Context Awareness. The results reflect the importance of context awareness for local and region-specific modifications. Our method InRetouch improves significantly on previous INR methods.