Difference between revisions of "AXIOM Beta TestsTODO"

From apertus wiki
Jump to: navigation, search
m
Line 7: Line 7:
#raw12 snapshots of real world scenes (outdoors, people, etc.) with high & low gain, long & short exposure times
#raw12 snapshots of real world scenes (outdoors, people, etc.) with high & low gain, long & short exposure times
#raw12 snapshots of real world scenes, overexposed and underexposed (to various degrees)
#raw12 snapshots of real world scenes, overexposed and underexposed (to various degrees)
#raw12 of the Hutch Color Target, with nicely bracketed range from blown out to dark lit by direct sun 5000k
#raw12 of the 3 color targets, with nicely bracketed range from blown out to dark lit by direct sun 5000k
#colorchecker passport zoomed in on the bottom half and defocused a bit (just the boxes area) - for FPN
#colorchecker passport zoomed in on the bottom half and defocused a bit (just the boxes area) - for FPN
#5 snapshots of an out of focus wall, taken at intervals of 10 minutes; repeat the experiment another day (to check FPN correlation)
#5 snapshots of an out of focus wall, taken at intervals of 10 minutes; repeat the experiment another day (to check FPN correlation)

Revision as of 00:03, 7 January 2016

  1. test flicker behavior of our light sources: 10 frames grey wall lit only by particular light with 0.1ms, 0.5ms and 1ms exposure times each
  2. raw12 snapshot of colorchecker passport with 2047 offset (regs 87/88) gain x1 and x2 - exposure so nothing clips
  3. repeat Hutch and IT8 charts - same framing for both Nikon and Apertus, please! (surroundings do matter, and so does the camera orientation)
  4. IT8 chart a little underexposed and a little overexposed on the gray scale, both Nikon and Apertus (0, -2 and +2 EV)
  5. darkframes, gains 1...4, exposures 1,2,5,10,25,50,100ms, 10 snaps at each setting, with entire camera packed in black cloth to make extra sure there cannot be any light leaking anywhere onto the sensor
  6. grey out of focus wall, gain x1,x2 with 100 frames of different exposure settings (listed below), making sure the last few frames are completely overexposed, on all channels; then run a script and upload the results (and the raw files as well, if possible)
  7. raw12 snapshots of real world scenes (outdoors, people, etc.) with high & low gain, long & short exposure times
  8. raw12 snapshots of real world scenes, overexposed and underexposed (to various degrees)
  9. raw12 of the 3 color targets, with nicely bracketed range from blown out to dark lit by direct sun 5000k
  10. colorchecker passport zoomed in on the bottom half and defocused a bit (just the boxes area) - for FPN
  11. 5 snapshots of an out of focus wall, taken at intervals of 10 minutes; repeat the experiment another day (to check FPN correlation)

For all samples:

  • Please set the black offset registers (87,88) to 2047. 1910 is too low (clipped shadows = data lost). To get a normal picture, use raw2dng --black=137.
  • Please save the register metadata block (on past samples, many files were mislabeled).

For the color charts:

  • Please use a static scene (same background in both Nikon and Apertus), and try to take the picture from the same location (they should have the same framing *after* cropping the Nikon image to APS-C).

Exposure settings for the gray frames:

exp_time reg: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 17, 18, 19, 21, 22, 24, 26, 28, 30, 33, 35, 38, 41, 44, 48, 52,
              56, 60, 65, 70, 75, 81, 88, 95, 102, 110, 119, 128, 138, 149, 161, 174, 188, 202, 218, 235, 254, 274, 296, 319,
              344, 371, 400, 432, 466, 503, 542, 585, 631, 681, 735, 793, 855, 922, 995, 1074, 1158, 1249, 1348, 1454, 1569,
              1692, 1826, 1970, 2125, 2292, 2473, 2668, 2878, 3105, 3349, 3613, 3898, 4205, 4537, 4894, 5280, 5696, 6145, 6629,
              7152, 7715, 8323, 8979, 9687.

To compute the exposure times, please write down the LVDS clock value used during the test.

Script to be run on those 100 frames:

raw2dng *.raw12
python2 raw12_stats.py