RealX3D is a real-capture 3D benchmark for restoration and reconstruction under in-the-wild degradations — low-light, blur, occlusion, reflection, etc.

DeMoE is the first all-in-one deblurring method that restores images with diverse blur types — global motion, local motion, low-light blur, and defocus.

Fast and efficient baselines for real-world low-light image enhancement. Process HD videos at 24 FPS. Fastest model in NTIRE challenges.

We propose the first approach that introduces language and explicit control into the image processing and editing pipeline. PixTalk is a vision-language multi-task image processing model, guided using text instructions. Our method is able to perform the most popular techniques in photography.

The best Bokeh rendering method using neural networks and a novel dataset. The neural model allows to control the apertur from f/16 to f/1.8

We present a novel analysis on compressing neural fields, with the focus on images. We also introduce Adaptive Neural Images (ANI), an efficient neural representation that enables adaptation to different inference or transmission requirements.

DarkIR In low-light conditions, you have noise and blur in the images, yet, previous methods cannot tackle dark noisy images and dark blurry using a single model. We propose the first approach for all-in-one low-light restoration including illumination, noisy and blur enhancement.

InstructIR takes as input a degraded image and a human-written instruction for how to improve that image. The (single) neural model performs all-in-one image restoration. We achieve state-of-the-art results on several restoration tasks including image denoising, deraining, deblurring, dehazing, and (low-light) image enhancement.

Implicit Neural Representations (INRs) are a novel paradigm for signal representation that have attracted considerable interest for image compression. We explore the critical yet overlooked limiting factors of INRs, such as computational cost, unstable performance, and robustness

We tackle image restoration directly in the RAW domain (yes, it is tricky and a bit crazy).

First, we propose a novel module that can be integrated into any neural ISP to capture the global context information from the full RAW images. Second, we propose an efficient and simple neural ISP that utilizes our proposed module.

NILUTs are neural representations of real 3D LUTs for controllable photo-realistic image enhancement and color manipulation. Moreover, a NILUT can be extended to incorporate multiple styles into a single network with the ability to blend styles implicitly.

Collaboration with Meta. This paper introduces a novel benchmark as part of the AIS 2024 Real-Time Image Super-Resolution (RTSR) Challenge, which aims to upscale compressed images from 540p to 4K resolution (4x factor) in real-time on commercial GPUs. The images are compressed using the modern AVIF codec, instead of JPEG.

Collaboration with DXOMARK. This paper reviews the NTIRE 2024 Portrait Quality Assessment Challenge, highlighting the proposed solutions and results. This challenge aims to obtain an efficient deep neural network capable of estimating the perceptual quality of real portrait photos.

New methods for RAW Super-Resolution could be essential in modern Image Signal Processing (ISP) pipelines, however, this problem is not as explored as in the RGB domain. The goal of this challenge is to upscale RAW Bayer images by 2x, considering unknown degradations such as noise and blur.

We advance RAW sensor images up-scaling (Super-Resolution). We explore diverse image degradations (e.g. Noise, Blur) to emulate a low-resolution RAW image, and we train a neural network to upsample it.

EBokehNet, an efficient state-of-the-art solution for Bokeh effect transformation and rendering. We can render Bokeh from all-in-focus images, or transform the Bokeh of one lens to the effect of another lens without harming the sharp foreground in the image.

The paper presents an exhaustive study of baseline methods for real-time image SR. The methods allow 60 FPS and even 120 FPS.

The paper gauges the state-of-the-art methods for real-time 4K upscaling using our new dataset and benchmark protocol. Ovr 100 participants joined the challenge and 15 teams proposed diverse solutions. The methods allow 60 FPS and even 120 FPS on regular GPUs.

We propose LPIENet, a lightweight network for perceptual image enhancement, with the focus on deploying it on smartphones. The model was tested for image denoising, deblurring, and HDR correction.

Super-resolution of compressed images using transformers. We use the Swin Transformer V2, to improve SwinIR for image super-resolution, and in particular, the compressed input scenario. Using this method we can tackle the major issues in training transformer vision models, such as training instability, resolution gaps between pre-training and fine-tuning.

This paper introduces the AIM 2022 Challenge on Reversed Image Signal Processing and RAW Reconstruction. We aim to recover raw sensor images from the corresponding RGBs without metadata and, by doing this, "reverse" the ISP transformation.

Hybrid model-based and data-driven approach for modelling ISPs using learnable dictionaries. We explore RAW image reconstruction and improve downstream tasks like RAW Image Denoising via raw data augmentation-synthesis.

We were one of the 1st attempts to use CLIP (Contrastive Language-Image Pre-Training) for training a neural network on a variety of art images and descriptions, being able to learn directly from raw descriptions about images, or if available, curated labels.