Single-Axis Multi-Parameter (SAMP) Camera
The Single-Axis Multi-Parameter (SAMP) Camera is a research project at MERL that is attempting to rectify the problems with existing surveillance cameras. Taking an image with a standard camera requires the photographer or the camera itself to choose the resolution, the aperture, the exposure time, focus parameters, frequency and polarization parameters (usually done with lens filters), and frame rate (if a video camera). The SAMP camera takes multiple images with different settings of all these parameters with one exposure. Using these multi-parameter images, typical image-processing operations become much easier and result in imagery of markedly better quality.
Background & Objective: In the context of computer vision, multiple images that are geometrically similar but radiometrically controlled are useful for many applications, such as high dynamic range imaging, focus/defocus analysis, multispectral imaging, high speed videography and high resolution imaging. We informally refer to this as "creating multiple copies of an image". Creating multiple copies of an image has been demonstrated in various contexts, but rarely more than three copies are created. This project addresses the question of how to make a larger number of copies within an optical system.
Technical Discussion: We have implemented a system using an optical splitting tree to generate up to eight copies of the scene. The internal nodes are optical elements like beam splitters, filters, and lenses, and the leaves are video sensors. Varying the optical elements allows us to capture at each virtual pixel multiple samples that vary not only in wavelength but also in other sampling parameters like focus, aperture, polarization, exposure, subpixel position, and frame time.
A wide variety of applications can be implemented using this system by simply reconfiguring its optical elements. We have implemented high dynamic range (HDR), multi-focus, high-speed, high-resolution video, and pulling a high-quality alpha matte and foreground from a video sequence.
| Technical Reports: | |
| Defocus Video Matting | |
Technology Areas:
Computer Vision
Sensor and Data Systems
Modification Date: September 12, 2007