Remanent Imaging (“RI”) is a video imaging paradigm that records motion blur to improve motion detection and analysis, at a massively reduced computational load and data storage footprint. It’s implemented by setting the camera exposure to match the time interval between frames. Darkfield illuminated RI subjects have highly visible trajectories that are continuous from image to image. Stacking image sequences recapitulates what a single long exposure photo would have recorded over the sequence duration. This recapitulation power is illustrated to the left with a stack of 75 images of 12 five-days-post-fertilization (“dpf”) zebrafish larvae spanning 20 seconds.
Because motion continuity is preserved from frame to frame, the precise fish motion is captured by the crest lines of the intensity profile of the image stack, and the rest of the image can be discarded, yielding massive data compression—10,000x in the above example. Most crucially, this compression is achieved without sacrificing time resolution: all C-starts ( the sharp snapping fish escape initiations that require millisecond sampling rates) are clearly visible on long exposures. Due to this data footprint, RI can realistically be used for lifetime recording.
Integrative by design, Remanent Imaging enables another major innovation: direct measure extraction from stacked image sequences at user-specified time resolution. In traditional image processing, videos are processed frame by frame to identify and track moving objects, build their trajectories, and then compile various measures. Because systematic error accumulates as a result of frame by frame compilation, further treatment is required for mitigation. RI sequences can be processed in this manner to track individual fish.
The PiscisTrack™ software architecture, illustrated to the left, is designed on the following foundational principles:
