| | 1 | == Large-Scale Masking Analysis == |
| | 2 | |
| | 3 | How can we convert the masks generated for each image into a resource which helps people to determine the filling factor of their particular science? Here are some thoughts: |
| | 4 | |
| | 5 | * we generate a mask image with each processed science exposure. These are also warped with the image pixels to skycells. The masks are compressed, and thus small (1% of the image data volume). |
| | 6 | |
| | 7 | * we ingest exposures into a DVO database, and that is made available to people for their studies. |
| | 8 | |
| | 9 | * anyone who cares only about a specific object or small area can always request the relevant masks from the postage stamp server |
| | 10 | |
| | 11 | * The total data volume of per-exposure masks generated to date is in the vicinity of 2TB. |
| | 12 | |
| | 13 | * Option 1) set up a distribution system for all masks to provide all masks to those who so desire. |
| | 14 | * this scheme, or others below, could use Mario Juric proposed mask compression tool to further reduce the volume. |
| | 15 | |
| | 16 | * Option 2) set up an analysis stage which generates binned mask images on various scales |
| | 17 | * this system could start with the warped skycell masks |
| | 18 | * these masks could be combined into a lower resolution set of super-skycells |
| | 19 | * a static-sky worth of masks at full resolution should require ~200GB per filter |
| | 20 | * binning by 5x5 (1 arcsec pixels) gets us to the 10GB range. |
| | 21 | * what scales should be generated? |
| | 22 | |
| | 23 | * Option 3) set up a system to rebin and distribute the masks, to reduce the volume of data I/O |
| | 24 | |
