Digital Exposure Experiments (2006)
Back in the days before smartphones had both reasonable cameras and plenty of CPU for processing data, I performed processing experiments with a Firewire video camera hooked up to a laptop.
In spite of the clunky appearance of the rig, the long term "exposures" made possible by this arrangement allowed for photographic opportunities that were simply not possible to capture with traditional hardware.
One of my objectives was to photograph the River Thames in a still state, which is obviously not possible through traditional means since the river is always flowing. In order to achieve this, I worked with 15 seconds of data collection (at 25 FPS this amounted to 325 individual frames per shot) followed by statistical analysis to eliminate the effects of the surface refraction.
Fig 1. Raw video frame
Fig 2. A simple averaging of frames produces a 'frozen over' look
Fig 3. More complex refracation cancellation gets us closer to a 'mirror finish'
Some of the interesting applications of this prototype were:
- Simulation of traditional long term exposures by averaging multiple video frames into a single frame.
- Noise cancellation in low light conditions by Kalman filtering of individual pixel offset components.
- Water ripple refraction cancellation through statistical analysis of multiple (300+) sequential frames.