Bookmark and Share

Software Topics

Using Computational Photography HDR Algorithms to Augment Low Cost Sensors in Computer Vision Applications
2 replies [Last post]
Eric Gregori
Eric Gregori's picture
Offline
Last seen: 3 years 34 weeks ago
Level 6: Enchanter
Joined: 2011-08-16
Points: 202

 

While researching computational photography I came across an interesting subset of algorithms for high dynamic range imaging.  By taking a series of over exposed and under exposed images and combining them using well know algorithms [Richard Szeliski  (http://szeliski.org/Book)], a final image of significantly better dynamic range can result. 

http://www.cis.rit.edu/fairchild/PDFs/PAP24.pdf

 

A company called  Almalence (http://almalence.com/doc/mobile-hdr-examples-deghosting/) has a product that can calculate an HDR result in 5 seconds on a 500Mhz ARM with a 5M pixel camera.  5 seconds is obviously too slow for real-time computer vision, but a 5M pixel camera is also a much higher resolution then is required by many CV applications.

 

Is anyone familiar with work being done on real-time HDR processing?

Eric Gregori
Eric Gregori's picture
Offline
Last seen: 3 years 34 weeks ago
Level 6: Enchanter
Joined: 2011-08-16
Points: 202

Another interesting paper covering computational photography and HDR:

 

http://www.cfar.umd.edu/~aagrawal/EG2006STAR_CompPhotog.pdf

 

 

Eric Gregori
Eric Gregori's picture
Offline
Last seen: 3 years 34 weeks ago
Level 6: Enchanter
Joined: 2011-08-16
Points: 202

I also found this interesing video demonstrating how HDR works.  The application the video describes is also interesting.  Magic Lantern is a "firmware upgrade" you can do to your Cannon camera to add functionality, including HDR.

 

http://youtu.be/Quahz5d2JFo