AXIOM Beta TestsTODO

From apertus wiki
Revision as of 15:34, 7 January 2016 by 193.170.138.1 (talk)
Jump to: navigation, search

1 TODO

  1. repeat Hutch and IT8 charts - same framing for both Nikon and Apertus, please! (surroundings do matter, and so does the camera orientation)
  2. IT8 chart a little underexposed and a little overexposed on the gray scale, both Nikon and Apertus (0, -2 and +2 EV)
  3. 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)
  4. raw12 snapshots of real world scenes (outdoors, people, etc.) with high & low gain, long & short exposure times
  5. raw12 snapshots of real world scenes, overexposed and underexposed (to various degrees)
  6. raw12 of the 3 color targets, with nicely bracketed range from blown out to dark lit by direct sun 5000k
  1. 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


2 Done

  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: http://files.apertus.org/AXIOM-Beta/snapshots/LightFlickerTest01/
  2. raw12 snapshot of colorchecker passport with 2047 offset (regs 87/88) gain x1 and x2 - exposure so nothing clips: http://files.apertus.org/AXIOM-Beta/snapshots/ColorChecker-Offset-test-07.01.2016/
  3. colorchecker passport zoomed in on the bottom half and defocused a bit (just the boxes area) - for FPN: http://files.apertus.org/AXIOM-Beta/snapshots/ColorCheckerOutofFocus-07.01.2016/
  4. 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: