05-11-2026, 06:33 AM
(This post was last modified: 05-11-2026, 06:39 AM by npcompl33t.)
Very interesting @KoenG, I like the idea. I checked and the MSI values show roughly the same values as you for the pages we have that data for. resampling all 16 Quire 4 faces gave B a +1.18 L* advantage, close to your reported +0.9.
I tested this against my clusters and path values, and there is a broad correlation. Pearson r = 0.426
Spearman ρ = 0.386
Looking at the chart, it seems like there is agreement at the beginning, and mostly at the end.
It seems like the page brightness is good at identifying the start of the ms, and the very end, but not so much at the pages in the middle, where many pages get approximately the same value. 27/30 is an odd outlier on your chart compared to mine, where it normally comes 1st or 2nd.
The way that a few at the beginning and end are pushed to the margins is interesting. Maybe that could be consistent with some sort of binding effect? where pages closest to the edges are impacted differently from ones in the middle?
I found a You are not allowed to view links. Register or Login to view. that discusses a method for calibrating RGB measurements from multispectral data. I'm going to see if I can use that to crosscheck the RGB values from the MSI data we do have, and should provide one additional check on these numbers.
I tested this against my clusters and path values, and there is a broad correlation. Pearson r = 0.426
Spearman ρ = 0.386
Looking at the chart, it seems like there is agreement at the beginning, and mostly at the end.
It seems like the page brightness is good at identifying the start of the ms, and the very end, but not so much at the pages in the middle, where many pages get approximately the same value. 27/30 is an odd outlier on your chart compared to mine, where it normally comes 1st or 2nd.
The way that a few at the beginning and end are pushed to the margins is interesting. Maybe that could be consistent with some sort of binding effect? where pages closest to the edges are impacted differently from ones in the middle?
I found a You are not allowed to view links. Register or Login to view. that discusses a method for calibrating RGB measurements from multispectral data. I'm going to see if I can use that to crosscheck the RGB values from the MSI data we do have, and should provide one additional check on these numbers.