computational photography

New Scientist reports on a project by Péter Schaffer and Djamila Aouada at the University of Luxembourg and Shishir Nagaraja at IIIT Delhi to protect people who photographically document abuses by oppressive regimes. After Nagaraja learnt that Burmese agents identified people who photographed protests using their positions in video footage, the research team set out to develop a method based on view synthesis. View synthesis takes images from multiple viewpoints and creates a new image from a different viewpoint. This provides "location privacy" for the photographers. The article describes this development as part of an "arms-race" between tools that enable regimes to automatically identify people in large sets of images and tools that promote privacy and security.

Clarification (Aug 1, 2011): During the Burmese event, only the photos made by the authorities were used to hunt down demonstrators who had cameras in their hands (i.e., photos made by demonstrators were not correlated to video footage).

contributions and corrections made by Péter Schaffer

Computational photography startup Lytro has announced its launch on its blog. Lytro plans to bring a camera based on a microlens array -- a dense array of imaging units each with its own lens -- to the consumer market. With this camera, it is possible to use software to change focus and shift viewpoint after the picture has been captured. According to the New York Times, Lytro's advance is to make the computationally intensive lightfield camera software efficient enough to run within a camera. Lytro has received $50 million in startup funding.

Related articles on Computer Vision Central:

• Columbia University developing a gigapixel single snapshot camera
• Multi-exposure protocol for high quality images
• Prediction: digital cameras to offer advisory and information role
• Free ebook on computational photography

A team of researchers has combined 3 million images from online sites like Flickr to build a 3D model of Rome's major landmarks. The feat was performed using computer vision software running on a single PC with commodity graphics hardware. The project was a collaboration between the University of North Carolina at Chapel Hill and ETZ-Zurich.

According to lead researcher Jan-Michael Frahm, “Our technique would be the equivalent of processing a stack of photos as high as the 828-meter Dubai Towers, using a single PC, versus the next best technique, which is the equivalent of processing a stack of photos 42 meters tall – as high as the ceiling of Notre Dame – using 62 PCs. This efficiency is essential if one is to fully utilize the billions of user-provided images continuously being uploaded to the Internet.”

More information is available in a press release and the ECCV 2010 paper.

Related article on Computer Vision Central:

• Improved Photosynth can model a city in hours

The open-source digital photography software platform, "Frankencamera," will be available as a free download for the Nokia N900. The software platform allows users to create novel camera capabilities and applications. The project is led by Dr. Marc Levoy (Stanford) together with Dr. Kari Pulli, a Nokia Fellow at the Nokia Research Center (NRC) Palo Alto. More information is available in a press release.

Related Article:
Stanford researchers build open-source camera.

Submitted by Boaz Super

In his Ph.D. research at the University of California San Diego (UCSD), Neel Joshi developed methods for enhancing a photo of a person's face using other images of that person. The work on "identity-specific priors" can perform deblurring, super-resolution, color-balancing, and exposure correction. Since people often take many pictures of their family and friends, good pictures of a person may often be available for improving a poor one. The work is reported in "Personal photo enhancement using example images," published in ACM Transactions on Graphics

Submitted by Boaz Super

U.K. company Imsense Ltd. announced the release of an improved version of its photo enhancement software imphoto. In addition to improved image quality, the software now has export options specific to the iPad's display size, and to standard Facebook and Twitter image sizes.

University of Washington (UW) and Cornell University researchers are developing PhotoCity, a system to create 3D renderings of buildings and cities from unstructured collections of 2D digital photos. To improve the quality of the 3D renderings, the researchers plan to use a social game for users to add images where needed. More information is available in a New York Times article.

The technical program for ICCP 2010 (International Conference on Computational Photography) is now available. This year, the conference will take place in the MIT campus in Cambridge, MA , from March 28 to March 30. This year on March 28, ICCP will also host a series of invited talks on the Foundations of Computational Photography.

The ICCP 2010 keynote speakers are:

Adobe Photoshop is celebrating its 20th anniversary with the release of an interesting video with the creators of Photoshop discussing the evolution of the product. The discussion in the video mentions Photoshop’s filters and plugins that enable special image effects.

Researchers at Tel-Aviv University (Israel) and Zhejiang University (P.R.China) have developed algorithms to calculate an image’s aesthetics based on photographers’ rules on composition (object placement, color, and shape). Results from their work are presented at the Eurographics 2010 conference in Norrköping, Sweden, in May, in a paper titled, "Optimizing Photo Composition" (Ligang G. Liu, Renjie J. Chen, Lior Wolf, and Daniel Cohen-Or).