Please join the Goergen Institute for Data Science and Artificial Intelligence (GIDS-AI) on Friday, April 25, 2025 for The Power of Iteration: Achieving Realistic Image Restoration Through Direct Inversion, a research talk with Mauricio Delbracio, Staff Research Scientist at Google Research. A Zoom option is available for virtual attendees.

Abstract: Despite the advancements in deep learning for image restoration, achieving realistic and faithful outputs remains a significant challenge due to the 'regression to the mean' effect. I will present Inversion by Direct Iteration (InDI), a supervised approach that overcomes this limitation by iteratively refining image quality through small, controlled steps. This iterative process, inspired by generative diffusion models, enables the generation of highly detailed and realistic restorations across various image degradation tasks. We demonstrate that InDI achieves state-of-the-art results in motion and out-of-focus deblurring, super-resolution, compression artifact removal, and denoising.  Following this, I will present our team's recent work on multi-modal guided restoration, which leverages information from diverse modalities (depth, edges, text) to enhance realism and authenticity in reconstructed images. To conclude, I will showcase some novel image processing capabilities implemented in Google Phones, showcasing the real-world impact of advanced restoration techniques.

Bio: Mauricio Delbracio is a Staff Research Scientist at Google Research. Prior to joining Google in 2019, he was an Assistant Professor at the Electrical Engineering Department of Universidad de la República in Uruguay. From 2013 to 2016 he was a postdoctoral researcher with the ECE Department at Duke University. He received his B.Sc. in Electrical Engineering from Universidad de la República in 2006 and his M.Sc. and Ph.D. in Applied Mathematics from École Normale Supérieure de Cachan in 2009 and 2013 respectively. His research focuses on image and signal processing, computational photography, inverse problems, and machine learning. Recently, he was a key contributor to the launch of Unblur, an image deblurring feature introduced with the Google Pixel 7/Pro, and Zoom Enhance, Google's first image-to-image diffusion model, optimized to run fully on-device. In 2016, Delbracio was awarded the Early Career Prize from the Society for Industrial and Applied Mathematics' Imaging Science Activity Group for his significant contributions to image processing.

View the recording of this research talk