Opening and Welcome

Keynote: Denoising as a Building Block for Imaging, Inverse Problems, and Machine Learning

• [Abstract]

Denoising is one of the oldest problems in imaging. In the last decade, the quality of denoising algorithms has reached phenomenal levels – almost as good as we can ever hope. There are thousands of papers on this topic, and their scope is vast and approaches so diverse that putting them in some order (as I will do) is both useful and challenging. I will describe what we can say about this general class of operators, and what makes them so special. I will argue that denoising is more important than ever; not simply as a process for removing noise, but especially now as a core engine and building block for much more complex tasks in imaging, inverse problems, and machine learning.

Peyman Milanfar

Principal Scientist/Director, Google Research

Best Paper Award

Winner Paper: Event-based Image Deblurring with Dynamic Motion Awareness

Presenter: Patricia Vitoria (Huawei Zurich Research Center)

Best Paper Honorable Mention

Winner Paper: Continuous Spectral Reconstruction from RGB Images via Implicit Neural Representation

Presenter: Ruikang Xu (University of Science and Technology of China)

Invited Talk: Combining Internal and External Constraints for Unrolling Shutter in Videos

Presenter: Eyal Naor (The Weizmann Institute of Science)

Invited Talk: Synthesizing Light Field Video from Monocular Video

Presenter: Shrisudhan Govindarajan (Indian Institute of Technology Madras)

☕ Tea Break

Challenge Winner: ToF Depth Completion Track

Presenter: Dewang Hou (Peking University)

Challenge Winner: Quad-Bayer Remosaic & RGBW Sensor Remosaic Tracks

Presenter: Lingchen Sun (OPPO Research Institute)

Invited Talk: Animation from Motion Blur or Rolling Shutter Distortion

Presenter: Zhihang Zhong (The University of Tokyo)

Invited Talk: Event-Based Fusion for Motion Deblurring with Cross-Modal Attention

Presenter: Lei Sun (ETH Zurich)

🍱 Break

Challenge Winner: RGBW Sensor Fusion Track

Presenter: DongJae Lee (Korea Advanced Institute of Science and Technology)

Challenge Runner-up: RGBW Sensor Remosaic Track

Presenter: Qiaosi Yi (East China Normal University)

Challenge Winner: UDC Image Restoration Track

Presenter: Yurui Zhu (University of Science and Technology of China)

Keynote: Recent Evolution on Image Sensing

• [Abstract]

The current growth of the image sensor market is mainly driven by demands for the mobile imaging market. The driving force is further expected to continue more than several years from now. Thus, it is worthy to consider what kinds of demands come from the mobile imaging market to understand what kinds of imaging technologies are required.

On the other hand, sensing applications, which utilize image sensor date not only for taking pictures but also for person identification, product inspection, obstacle avoidance and so on, are emerging. They seem to growth in industry and automotive markets at first but expected in mobile market soon. In sensing applications, non-RGB image sensors are utilized as well as RGB image sensors.

In this talk, recent evolutions of image sensors for both RGB and other than RGB are introduced. First, RGB imaging technologies which have been implemented in recent image sensor products for mobile market are introduced. Second, some new-generation image sensors which are to acquire information other than RGB are introduced. Also, expectations that such new-generation image sensors are promising for mobile imaging applications in the future are explained by introducing some recent researches in vision-based imaging area.

Tomoo Mitsunaga

Manager, Sony Europe B.V.

Keynote: Robust Imaging Color and the Dual-Purpose ISP

• [Abstract]

In this talk, I will discuss recent work targeting imaging strategies to allow robust color imaging. In particular, I will overview capture-time algorithms that allows processed images to be unprocessed to correct potential errors that affect image color. I will also discuss the idea of a "dual-purpose" image signal processor (ISP) that is suitable for both photographic and scientific tasks.

Michael S. Brown

Professor, York University

Keynote: Computer Vision, One Photon at a Time

• [Abstract]

Single-photon avalanche diodes (SPADs) are an emerging sensor technology that promises single-photon sensitivity and the ability to time-tag photons with picosecond precision. In the last few years, we have witnessed a single-photon revolution, resulting for the first time in multi-megapixel SPAD arrays, which allow capturing sufficient spatial details needed for mainstream computer vision applications, including object detection and tracking, text recognition, pose recognition, and SLAM. This raises the following questions: Are SPADs ready to be used beyond niche scientific applications, as general-purpose computer vision sensors? Can they be deployed outside the lab in-the-wild? If so, what are their benefits over conventional cameras?

Mohit Gupta

Associate Professor, University of Wisconsin-Madison

Closing