A Look at Information Loss in Diffraction-Limited Data Compression

In the last twenty years, advances in new microscopy techniques have made it possible to examine the location of a single molecule with nanometer-scale precision and reconstruct the world of biology from the level of a single molecule to the story of a cell (and potentially an entire organism). Professionals in biophysics rely on precise measurements to make accurate claims, especially for diffraction-limited particles of interest, as the slightest variation in the data could alter the analysis. Most institutions are rapidly moving towards open data policies where data must be archived and made available at publication time. In addition, data retention policies are now commonplace. For TB-scale data sets, these requirements become cost-prohibitive, so practical archival of data files necessitates their compression. Until now, little research has been done on the feasibility of scientific data compression and whether the relevant information is accurately preserved therein. We examined the fidelity of four modern codecs in their ability to preserve positional data in synthetically generated images. Our results suggest that 80% compression is attainable for single images, although pixel-locking effects may prove challenging to overcome.