I am interested to know how people managed to get a good measurements for a set of pastels.
The best fit I can get for the “Generate cone catch model using chart” function is shown in the pic below. What spectro setup did everyone use? What were your integration times? I am using natural sunlight as I did for the chart photos. I used an integration time of 30ms and each value is an average of 100 measurements. I find that if I lower the integration time the fit is worse but the measurements are less easily saturated.
I am running out of time to get this to work. Any advice would be greatly appreciated.
Hi Silas, I’m in the same boat, and will be trying to do this, possibly next week. I’ll let you know if we make any progress.
The spectral measurements of the pastel reflectance needs to be of very high quality for this method to work well. Essentially the camera and spectrometer need to be receiving the same reflected light from the pastels.
Experience has taught me that reflectance can\’t be measured reliably with light from a probe (because distance to the surface is too critical), and can\’t be measured with an artificial light source (due to fluorescence issues, metamerism etc..) An integration sphere is also not ideal because the camera and spectrometer will both be receiving reflected light from a fixed angle.
But that actually results in a fairly simple setup. You need:
- Spectrometer, ideally with collimating lens, but failing that a standard probe. Sensitivity from about 310-720nm.
- A diffuse white standard (e.g. PTFE spectralon or a compressed pile of barium sulphate).
- Pastels (e.g. set of 64, remove any really fluorescent ones). Clean the pastels\’ top surfaces by rubbing them against a flat piece of paper.
- A flat surface to place everything on outside. The surface of the pastels must all be uniformly flat, and in the same plane as the white standard.
- Sunshine (try to use a day where there are no clouds in the sky at all, with the sun as high as possible – near mid-day).
Then follow these steps:
- Set the spectrometer white point using the white standard, and it\’s black point as you usually would. Always measure surfaces from e.g. 45-90 degrees from the surface (vertical/elevation) and ~80 degrees (horizontally/azimuth) relative to the sun. Ensuring the probe/lens doesn\’t create shadows on the surface you\’re measuring, or reflect light back on to the surface from the sun, and that the probe/lens surface itself is not illuminated by the sun (causing lens flare). This is all pretty critical! The distance from the probe to the surface is not hugely critical now that we\’re using a stable, directional light source (the sun), as long as you know the patch size being measured. The rest of the sky will still add light to the surface, so for best results keep the pastels as openly illuminated as possible, or the illumination between camera and spectrometer as uniform as possible. Alternatively set up the apparatus in a black box that only allows the direct sunlight in from the side (probably overkill).
- Measure the reflectance of one row of standards as fast as possible. At the end of the row, measure the white standard and black point again to check there has been no loss of either white or black point. If it doesn\’t look perfect repeat the row again. Carry on working through the chart row-by-row, test-measure-test. Be aware that the sun is constantly moving (and the atmosphere slightly altering the lighting), so assume the calibration only holds for a few seconds.
- Photograph the pastels and white standard without moving anything, from the same direction as the spectrometer measurements were made. Using RAW photos, exposure bracketing to get a good exposure etc… Create an mspec image as usual, using the same white standard as the spectrometer.
- Check out the format of existing colour chart files and match this with your own. Give it a filename without special characters.
This should give you the best possible agreement between camera and spectrometer. In practice, once measured (and pastels kept out of the light), you should be able to calibrate more cameras without going through the spectral measurements again. It should be fairly robust to slight variations in terrestrial daylight illumination too.