Natural Color from ChemCam Data Options

[jccwrt]

I had an idea for ChemCam dataviz that I've been developing on and off for the last month, finally seem to have the pipeline mostly working now.

One thing that interested me is seeing the color of the small plasma plumes generated during the sampling process. From some press release images by the ChemCam team we know that the plume colors for basaltic materials are variable, so I was wondering exactly how variable they were. Since ChemCam's spectral range entirely covers the visible region of the spectrum, you can use the data to reconstruct its observations in natural color. I decided it would be interesting to try and reconstruct plasma colors from the "cleaned, corrected spectrum" (CCS) dataset. This dataset theoretically isolates the plasma emission, so it provide the best possible reconstruction of plasma color.

Conceptually the project is relatively simple - grab a spectrum, convert it to a CIE colorspace, and then plot the resulting color. Fortunately a lot of the hard parts were already handled for me - the SciPython blog had a useful post about converting a spectrum to a CIE colorspace, and the SpectRes package had a great function for resampling the ChemCam spectrum for use with the colour_system class outlined in the SciPython blog post. Most of the development time was spent figuring out how to visualize and export the data, which I'm finally glad to have finished. Still working out the best way to combine the resulting color plots with the RMI image, but this is probably well beyond my current Python programming skills, so I'll probably end up resorting to Photoshop.

Here's an example of a 9 x 1 raster of shots taken on the Quela drillhole on Sol 1466:



And here's the corresponding RMI image mosaic showing shot locations on the drillhole. Note that this was a 10x1 observation, my guess is that the first spot was not processed to the CCS level because it hit the tailings and not the drill wall. Also note that the second shot is probably clipping the tailings pile, which is probably why the bottom color in the plasma visualization is so wildly variable.



The colors tend to dance around but are generally purplish. Since these spectra are collapsing roughly 3400 observations down into a 3-dimensional parameter, getting to a particular color is a non-unique solution. However, from looking at the spectrum a bit, bluish colors are probably dominated by calcium line emission in the violet, red-purple colors are probably from increased oxygen and potassium line emissions in the orange/red, and the lightness of the color is probably influenced by iron line emission in the green and sodium line emission in the yellow. My plan is to go through and calculate these colors for the various drillholes where CCS data is available, then check out other samples of interest. If you all have ideas of good spots to check out let me know!

[fredk]

Very cool. So do you present the final images in sRGB?

The plasma was imaged by MAHLI on sol 687 - here's an example: https://mars.nasa.gov/msl-raw-images/msss/0...993M60_DXXX.jpg

I don't know if the raw images are saturated or not. If not, there should be some colour information there for comparison with what you get from the spectra.

[jccwrt]

Very cool. So do you present the final images in sRGB?

The plasma was imaged by MAHLI on sol 687 - here's an example: https://mars.nasa.gov/msl-raw-images/msss/0...993M60_DXXX.jpg

I don't know if the raw images are saturated or not. If not, there should be some colour information there for comparison with what you get from the spectra.

Yep, the final images use the sRGB space!

Thanks for the extra point of comparison, by the way. I think the MastCam images are saturated (at least in the center of the laser spot), but the edge of the spot does have a pinkish color which seems to match pretty well with the colors I"m getting from the converted spectra.

[jccwrt]

First attempt at packaging everything together...unfortunately haven't figured out how to export an animated gif from Photoshop yet without relying on a more limited colorspace...