In 2011 the first commercial light field camera hit the market: the Lytro F01. A consumer toy, it offered the ability to refocus images post capture, and not much else. The file formats were proprietary, and the software did not expose raw sensor data or 4D light fields. Options were limited for those interested in exploring the properties of the underlying light field images.
Light fields capture 2D texture, 3D geometry, and higher order effects like refraction, reflection, speculars and subsurface scattering. First conceived for efficiently rendering complex scenes, light fields have proven useful over a broad range of imaging and vision applications, often yielding simpler and more robust solutions than their 2D counterparts. But to benefit from the light field we must work directly with its 4D structure.
The light field toolbox was created to address this need. Currently in its fourth release, it directly exposes raw sensor data and decoded 4D light fields. It works with both models of Lytro camera available today, as well as camera arrays and gantries, including the light fields from the Stanford light field archive lightfield.stanford.edu.
Importantly for roboticists, the Lytro cameras can be calibrated and their imagery rectified, creating useful 4D images that linearly map pixels to 4D rays in space. Frequency domain planar and volumetric focus filters, and a spatial domain planar focus filter demonstrate some of the capabilities of this rectified 4D structure. Future plans include calibration and rectification for the Raytrix line of light field cameras, improved calibration of 2ndgeneration Lytro cameras, and calibration of array and mirror style light field cameras.
Light Field Toolbox for Matlab: http://www.mathworks.com/matlabcentral/fileexchange/49683
Google+ Community: https://plus.google.com/communities/114934462920613225440
Donald Dansereau: https://profiles.stanford.edu/donald-dansereau
Australian Centre for Robotic Vision
2 George Street Brisbane, 4001
+61 7 3138 7549