Another post giving the math for converting RGB to greyscale is found here.
It seems that Luminosity, Lightness and Averaging are three methods which are used often. I am looking for someone who can find which of these three methods Photoshop 6 and CorelDRAW x5 uses to convert RGB to greyscale. (CorelDRAW under “autoadjust” and

In photoshop 6, thresholding a black and white image just doesn’t really help (in my opinion) because there is too much loss of the greys.Below is an image (used many times before) of SP-D (not my AFM images, someone elses available online) which I have manipulated in photoshop using image/adjust/levels, and image/adjust/autolevels, and image/adjust/autocontrast, and image/adjust/contrast in various ways. I will plot my favorite arm (upper left) which has three peaks between the N terminus center high point and the CRD, having probably appeared more times than any other configuration, and the pic is also biased to my assumption from previous LUT plots, to have three different heights for those three peaks, and also different widths. I will post plots in for that same arm, in the same order below. Top left, image as used previously (test 1) provides a great plot.

Test 1, image as used previously, not very much adjusted (i cant remember but likely some contrast increased and brightness reduced after the original conversion from color) but done in the same manner as other SP-D dodecamers that have LUT plots (see countless previous posts).
Confirming, and reconfirming previous LUT plots, there is an N terminus peak, three intermediate peaks (in this case the third peak from the N terminus is pretty small) and the bumpy peak of CRD (owing to its distribution of the three CRD domains in a large bulb like end.
Little adjustments, same peak, so that points to the small effect that the several different algorithms for changing RGB to greyscale may not matter as much as I thought. Here is a link to another article on converting color to greyscale. and yet another on terminology and greyscale (ok, so i need to remember to use grayscale not greyscale) conversion.
While i know to little about this to have an opinion, my basic understanding is that most of the conversion software attempts to mimic what the human ee sees, rather than wht the RGB actually convert to.. in terms of — Gray = (Red + Green + Blue) / 3 — which for science, i.e. looking over micrographs) probably needs to be as free of attempts to increase or decrease the “RGB” values to fit human perception. SO, that said. perhaps less manipulations, fewer algorithms might be the best approach.
I have used RGB conversion in photoshop, in coreldraw, and desaturation to convert RGB to grayscale (yep now gray with an “a”).

After reading (rather stumbling through) some manuscripts and blogs about dozens of RGB to grayscale equations … my decision is to go back to basic grayscale = R+G+B/3 rather than use all the fancy formulas that make the grayscale look more pleasing to the human eye and automatically adjusting the blue and greens to make things look “nice”. Then I can adjust the contrast to what pleases my eye, emphasizes what I am looking for, and brings out the structural detail in my images, manually, and define it as such. To find the codes that were used in some programs is totally impossible, and if found are probably just as biased a conversion effected by my “hand made manipulation” of the images. If the conversion from RGB to grayscale is standard and the adjustments, tweeks are mine I can just comment on them in a line of text in the materials and methods… ha ha

Who knew that was (could be) so complicated. Hahaha…

FOR ME…. my eye is still the best tool to convert and desaturate and change the brightness and contrast.  It might very well be biased, but at least it is not brainless.

It is easy to see that there is a definite width variation in at least three different places. Center N terminus peak is not the widest, though it is the tallest “Lightness” or “luminance” peak. The CRD (four outside areas are the largest in terms of width, but not the tallest but next tallest and it is the greatest in square nm (in this image, each being about 200 nm² while the N terminus is about 100 nm²).  I guess at some point someone needs to do area under the curves for both width and height and get a total nm for each of the regions.  It seems that there are three peaks and also a little blip for the area just before the CRD (presumably the coiled coil neck region, that means four peaks between the N terminus and CRD.  And sometimes it looks like 5.

Nm grid overlying area of the N terminus and CRD shown below.  The nm of the largest peak between (and next to N terminus) and CRD is something about half the size of the N terminus nm².

I found this article by Martin van Eijk et al which had rotary shadowed images of surfactant protein D.  One of these (which might be a dodecamer…with perhaps one of the trimers (on the left of the image) that is a little unwound? perhaps, or it also may be a structure with more than the four trimers of a dodecamer) but it has a small center area which presumably might indicate a different type of N terminus binding than human SP-D.  I would like to find AFM images of porcine SP-D to compare since this image (shadowed) and AFM produce such different views of SP-D that they are not comparable.

It will be fun to count the arms on the fuzzyballs in this image.  It appears that the porcine SP-D dodecamer (which i have measured as the diameter of a circle (dotted red line) that intersects the CRD) is just over 100nm, actually about 110nm according to their bar marker (red below). The ratio of CRD to length and width of the arms of the collagen-like and N terminus domains “looks” at first glance to be similar to hSP-D.

The dark area here in a fuzzy ball (not my AFM image, someone elses) shows a consistent 6-8 nm dark center…. between the N terminus peaks.  Both arms (as best as I was able to stretch them from the CRD across the fuzzyball (which may not be at all accurate, and i think is NOT accurate but used for now) the center is vacant.  or at least it has a valley in the LUT tables that is nearly as low as the valley before peak 1, 2, and 3 begin on the left and right halves of the molecule.  I think it is more likely that two trimers bend into a U shape and are connected somehow side to side.  I will continue to measure the arms as separate entities.

On the surface this might seem like a thankless task but i could be a valuable lesson in how collectins link together to make multimers.  Side to side links at the N termini in SP-A and MBP and other c-type lectins makes it very likely, more likely in fact because of images like that below, than any other arrangement.

Measuring here the second arm of a 14-arm SP-D fuzzyball to determine whether the pattern of LUT plots made from each arm match up the plots for the dodecamers. There is clear pattern to the latter, and seems from two LUT plot sets to be true for the fuzzyball. The problem arises from the discrepancy between the dodecamers and the central area of the fuzzyball which seems to be more like a ring than a huge accumulation of N termini.

Well, it seems pretty clear that the LUT plots gathered in ImageJ really are not affected much by the manipulation and scrambling around to make them readable in a program that doesn’t have a lot of line drawing flexibility.  I am convinced that the smallest amount of care will produce quite repeatable results in the LUT plots.  These images are photoshopped, cut, centered, rotated, saved and resaved as RGB or BW images, and the plots don’t really get too different, one to the other. This is good news i guess, hopefully it will be useful to someone studying SP-D
ON THIS HALF arm of a multimeric molecule (n=14) it still looks like the N terminus (left and a portion of the N terminus ring in fuzzyballs) will be the tallest peak (whitest – brightest – most luminous – lightest —still looking for that word to describe the digital whiteness) at or around 240 on the scale of 1-256. After that a close by and very consistent tall but not too wide peak, then possibly 2 or 3 small peaks and finally the carbohydrate recognition domain (not as high a peak as found for this domain – (whitest – brightest – most luminous – lightest level for the multi-lobed CRD).  Bottom two images of SP-D are photoshop images and NOT REAL., Modeled however after the top three images are of one arm which came from a single fuzzyball. (fuzzyball 1 arm 1)