FlexISP: A Flexible Camera Image Processing Framework

We replace the traditional imaging pipeline with an end-to-end optimization framework that is aware of the camera model, enforces natural-image priors, and directly solves for a particular output representation (e.g., YUV, DCT), jointly accounting for common image processing steps like demosaicking, denoising, superresolution, achieving superior results.

Felix Heide, University of British Columbia
Markus Steinberger, NVIDIA
Yun-Ta Tsai, NVIDIA
Nasa Rouf, University of British Columbia
Dawid Pajak, NVIDIA
Dikpal Reddy, NVIDIA
Orazio Gallo, NVIDIA
Jing Liu, University of California
Wolfgang Heidrich, University of British Columbia
Karen Egiazarian, NVIDIA
Jan Kautz, NVIDIA
Kari Pulli, NVIDIA

Fast Burst Images Denoising

This paper presents a fast denoising method that produces a clean, ghost-free image from a burst of noisy images.

Ziwei Liu, The Chinese University of Hong Kong
Lu Yuan, Microsoft Research Asia
Xiaoou Tang, The Chinese University of Hong Kong
Matt Uyttendaele, Microsoft Research Technologies
Jian Sun, Microsoft Research Asia

Spatial-spectral Encoded Compressive Hyperspectral Imaging

This paper presents a spatial-spectral encoded compressive hyperspectral (HS) images acquisition scheme (SSCSI) for high-resolution snapshot HS imaging by analyzing and evaluating sparse representations of natural HS images, robust 3D HS images reconstruction from 2D coded projections, and additional applications include 3D HS images compression and denoising.

Xing Lin, Tsinghua University
Yebin Liu, Tsinghua University
Jiamin Wu, Tsinghua University
Qionghai Dai, Tsinghua University

Mirror Mirror: Crowdsourcing Better Portraits

We describe a crowdsourcing method for providing feedback on portrait expressions, and for selecting the most attractive expressions from large video/photo collections.

Junyan Zhu, University of California
Aseem Agarwala, Adobe Systems, Inc.
Alexei Efros, University of California
Eli Shechtman, Adobe Systems Inc.
Jue Wang, Adobe Systems Inc.