January 30, 2014
Following on from Matt’s post about the difficulty of photographing big specimens without distortion, I thought I’d have a play with our best Sauroposeidon C8 photo, which I think is this one:
(That’s been the basis for classic SV-POW! posts such as Your neck is pathetic and Darren’s new indeterminate Wealden maniraptoran is inadequate.)
I was motivated by Andy Farke’s comment:
Another–and perhaps more important–area where surface models excel is when you can remove colors on the original specimen that wash out relevant details…I bet this is probably the case for the example vertebra of Sauroposeidon. How many fossae and foramina just don’t show up well on the photos above?
Andy was talking about completely colourless 3d surface models, in which the 3d shape allows a render to make shadows that bring out the subtle shapes. But it made me wonder whether we could get anywhere just by washing out the most prevalent colour in the photo.
I started by doing a big, fat Gaussian blur on a duplicate layer — 500 pixels in each direction — and sampling the colour in the middle, to get a rough-and-ready average. (There may be a better way — please shout if you know one.) That average colour was#7e6b2f. I used it to run Colour To Alpha on another duplicate of the original layer, so that we’d be left with only residual colours. Here’s the result:
I’m in two minds about this. It may be informative, but it sure is ugly. To compromise, I reinstated the original layer underneath this mostly-transparent one, and turned its opacity down to 75%. Here’s the result — a nice compromise:
Of course, there are endless other approaches you can take — that’s the blessing and the curse of image-editing programs like GIMP. For example, here’s what I got doing a simple Colours → Auto → White Balance:
I’m not sure that isn’t the best of the bunch, in terms of informativeness.
I also tried something else — not amazingly successfully, but I think it’s worth seeing. Since the two photos that Matt showed in the previous post were evidently taken from somewhat different angles, I thought I’d have a go at compositing them into a red-cyan anaglyph. Because the variation in camera position is mostly dorsoventral rather than anteroposterior, the vert has to be pointed upwards for the two eyes to see the two versions from different horizontal points. Here’s the best I could do:
I would say this is of some value; but it’s nowhere near as good as, for example, the anaglyph of Cervical S of the Archbishop. I could sit and look at that one all day. The problems with this one arise for three reasons.
First, I had to reduce both parts of the Sauroposeidon anaglyph to monochrome (since one was already in that form), so all colour information was lost.
Second, I had to scale the high-resolution picture to the same size as the lower-resolution one, throwing away more detail.
Finally, and most important, the two photos were not taken with the intention that they should be used to make an anaglyph. To work well, this has to be done with the images taken under the same lighting conditions, at the same distance from the specimen, from perspectives differing by about the distance between the pupils of the viewer, and with the camera-position difference being perfectly in the plane of the specimen. Needless to say, none of these conditions was met in this case, so it’s actually quite impressive that it works as well as it does.
We have a lot of options for illustrating specimens these days. Postage-stamp-sized greyscale photos really don’t cut it any more.
In his post on Vicki’s new book Broken Bones, Matt told us his twelve-step process for producing stippled illustrations like this one of a crushed skull, which became the cover image of the book:
As soon as I saw that, I found myself thinking that it would look nice with some shading of the bone. Of course the existing stippling is a perfect guide to how dense the shading should be at each point, so I figured there had to be an easy way to do this automatically. There is, and this is what I whipped up in five minutes:
Here’s how I did it.
- I loaded Matt’s image into the GIMP, my image editor of choice.
- For some reason the crucial next step doesn’t work with greyscale images, so I converted it to RGB (Image → Mode → RGB)
- I removed the white background, leaving it transparent (Colours → Colour to Alpha… and click OK on the default colour, white)
- I added a new all-white background layer.
- I duplicated the skull layer, and named it “shading”
- I blurred the shading layer by 50 pixels (Filters → Blur → Gaussian Blur…, set the blur radius to 50 pixels and hit OK.) That gives you the shading you want, but it smudges out past the outline of the skull, hence the last two steps:
- I went back to the skull layer, and using the Fuzzy Select Tool (magic wand) selected the contiguous transparent area outside the skull parts.
- I went back to the shading layer and cut the selected area, leaving only that shading that’s inside the boundary of the skull.
As always with Gimp tutorials, it takes about ten times as long to explain as to actually do.
When I showed this to Matt, I rather immodestly said I was “super-happy with it”. Matt said he was “super-happy with the idea, but only regular happy with this specific execution”. He felt that the blurring was too strong, and that it should be backed off by 30-40%. So I made a new shading layer in the same way as above, but this time blurring by only 30 pixels. Here’s the resulting image:
It’s quite a subtle difference, but clear if you flip back and forth between the images (which you can most easily do by putting them in adjacent tabs of your browser). Personally, I think I prefer the 50-pixel version, since I think the shading clings rather too closely to the lines in this one, but YMMV.
Since I had both blur layers right there in the image, I thought it might be interesting to see how they look together. Here’s the result:
I’m actually rather fond of this version, but it’s a long way from the crisp, clinical feel of the original.
You can thicken up the shading by duplicating one or both of the shading layers as many times as you wish (or or course thin it out by sliding down the opacity level). Its also easy to make the shading coloured: just use Colours → Levels, select the individual colour channels, and bring up their bottom levels to taste.
Putting all that together, here’s one I made with very dense, yellowish (bone-coloured) shading. I did it starting with the 50-pixel shading layer, upping the red output level to 200 and the green to 150, then duplicating that layer, and reducing the 30-pixel shading layer to 50% opacity.
You can play for hours with all these sliders, tweaking as you wish, thanks to the magic of layers. It’s well worth investing a bit of time to learn some of the capabilities of a program like GIMP. Matt and I are very far from wizards, but we have at least got a bit past just using it to cut out backgrounds, and it opens up possibilities.
December 18, 2013
When we last left my better half, Dr. Vicki Wedel, she was helping to identify a Jane Doe who had been dead for 37 years by counting growth rings in the woman’s teeth. That case nicely illustrated Vicki’s overriding interest: to advance forensic anthropology by developing new methods and refining existing ones. To that end, for the past few years she has been working with her PhD advisor, Dr. Alison Galloway at UC Santa Cruz, to revise and update Alison’s 1999 book, Broken Bones: Anthropological Analysis of Blunt Force Trauma. The revised and much expanded (504 pages vs 371) second edition, co-edited by Vicki and Alison, came out Monday (Amazon, Amazon.co.uk).
You can read the whole table of contents on Amazon (click to look INSIDE!), but the short short version is that the book has three major sections. The first covers the science and practice of trauma analysis (pp. 5-130), and the second classifies hundreds of common fractures throughout the skeleton, with illustrations (pp. 133-313). The chapters in these sections were all written by Vicki, Alison, and another of Alison’s former students, Dr. Lauren Zephro, solo and in varying combinations. Lauren, whom I always think of as “the Amazon cop”, is a 6-foot blackbelt forensic anthropologist for the Santa Cruz County Sheriff’s Office. If push came to shove I have no doubt that she could beat me to death with her bare hands and then produce a technical analysis of my corpse.
The final section (pp. 314-410) consists of nine case studies contributed by forensic anthropologists, pathologists, medical examiners, forensic and medical artists, and a DoD casualty analyst, based across the Anglophone world from Hawaii to Scotland. There’s some grim stuff in there: trauma to the homeless and elderly, from intimate partner violence, and from child abuse. It’s gut-churningly awful that the defenseless suffer from bone-breaking violence; it’s always been amazing to me that people like Vicki, Alison, Lauren, and the other contributors have both the courage to face these horrors and the technical chops to make the unspeakable solvable.
Beyond that unavoidable darkness, if you’re interested in the many, varied, and often just plain weird ways in which people die, the book is a treasure trove. There’s an elderly woman lying on her deathbed for six years, slowly turning into a natural mummy… (wait for it) …while her daughter went on living in the same house. There’s a classification of plane crashes with a description of what human remains will be found and over what area. There are people hit by trains; the funniest line in this very serious book is the deadpan and unsurprising, “The typical pedestrian hit by a train is male and often highly intoxicated.”
That’s from the top of page 122. At the bottom of the same page is my one contribution to the book, which also appears as the cover art (yeah, nepotism, whatcha gonna do). There’s a story behind this. This guy–yes, male, dunno if he was intoxicated–was hit by a train and his head was sheared in half, with the somewhat fractured but mostly intact facial skeleton separated by a lot of missing bone from the occipital region. With no way to obtain the deceased guy’s permission to use his mortal remains in the book, Alison and Vicki didn’t feel comfortable including their photos, so I spent a weekend bashing out a technical drawing for them to use instead. That reawakened my interest in pen-and-ink work and led to the dödö pöst.
I should say two things right here: first, that yes, I am hijacking the rest of the post to talk about myself. (Is anyone really surprised? I thought not.) Second, that I have had no training and possibly my stippling violates Art Rules or best practice guidelines of which I am ignorant. But I hope it also illustrates what can be achieved in a couple of days, with about $15 worth of supplies, by a guy whose only rule is “möre döts”.
So anyway, if you’re curious, here’s the method I use for my pen-and-ink illustrations:
- Get a decent-sized photo of the object to be drawn. I usually roll with $2 8x10s from MalWart, in this case one for each half of the skull.
- Tape the photo down to your work surface. I have a large, incredibly hard, perfectly smooth cutting board that I use for this, but in a pinch you could use just about anything, including just a larger piece of paper. Cardboard off the backs of desk calendars is nice.
- Over the photo, tape down a piece of tracing paper.
- Lightly trace the outline of the photo and all the major details in pencil.
- Once you’ve gone as far as you can with that, peel up one side of the tracing paper, unstick the photo from the work surface, and remove it. Stick the tracing paper back down the work surface.
- Using the uncovered photo as a reference, pencil in any other salient details by eye. Also contour lines for shadows. All of the pencil lines are going to be erased later, so don’t be shy.
- Whenever you decide you’re done with the pencil, get a good pen and start tracing, directly over the pencil lines. I tend not to be too persnicketty about my tools but decent pens are a real help here. For these recent works, I picked up a three-pack of beige-tubed Micron pens for $7 (this set).
- In all of the following pen-related steps, be careful to keep your big stupid hand and arm off the wet ink you just laid down–one careless smear can ruin a few hours’ work. Having a work surface that you can rotate is nice, so your pen hand can approach the drawing from any angle. Anytime I have to lay my hand on the drawing, I put down a piece of clean scrap paper first. Even if the underlying ink is dry, it just feels like a smart precaution.
- Once the lines are on, add döts to taste. With a little experimentation, you can get patterns of dots to not only indicate light and dark but also suggest textures. Different pen tips and amounts of pressure will yield dots of different sizes, which can also be useful. Dense, overlapping dots can produce an effect similar to scratchboard. BTW, sometimes I do “gear down” and place each dot with thought and care, but in the dense sections I just rat-a-tat-tat like a Lilliputian jackhammer. Try different speeds and see what you can tolerate.
- When you’re done dötting, at least to a first approximation, and you’re dead certain the ink is all dry, get a decent eraser and erase all of the pencil lines. I used one of those clicky mechanical erasers because it was cheap and soft enough to not tear up the paper.
- Re-ink any lines lightened by the erasing. I find that the döts are usually unaffected, but lines are often knocked down a bit by the eraser work. I suppose it would be cleaner to just draw natively in pen, with no prior pencilling and therefore no erasing, but the few times I’ve tried it, it hasn’t gone well. YMMV. If you’re drawing a 3D solid, this is a good time to employ an old illustrator’s trick, which is to make the bottom outline heavier and darker than the rest, to subtly convey a sense of weight.
- Scan, touch up as needed in GIMP, post to blog, bask in self-admiration.
In this case I had a few more steps, which consisted of making variants of the drawing and test-driving them by Vicki and Alison so they could pick their favorites.
This is just embarrassing: after scanning the two drawings and doing a little touch-up, I just scooted them together until they looked like a skull. The problem is that the occiput is nowhere near anatomical position. See that flange of bone above the ear-hole, pointed down and right at a 45-degree angle? That’s the back end of the zygomatic arch, and it should be aimed at the forward stump of the arch, which is just down and back from the eye socket.
Here’s the B version, where I was working entirely off of the zygomatic arch ends, and trying to get the skull into anatomical position. Scientifically this is probably the best variant I produced (I’m not claiming it’s the best possible), but aesthetically it’s a little crowded.
I’ll spare you versions C-E, all of which just scooted the back end of the skull around in an attempt to find a balance between scientific and aesthetic concerns. Here’s the winning F version, which got used for the figure, and became the seed variant for the cover.
For the cover, we tried a lot of things, including the white skull on a black background, and one that was simply inverted from the figure. By this point the publisher had sent Vicki some test versions of the cover, and I thought it would be cool if the drawing was in the same color as the cover text, so I sampled that color from the publisher’s sample cover image and applied it to all of the drawn bits. They knocked it down a few tones for the printed version, so happily it’s not this garish.
Incidentally, I had never tried to replace a bunch of discontinuous areas of the same color with another color in GIMP, so I had to look it up. The two key steps are Select > By color, with the threshold set to zero (or not, if you want to grab a bunch of related colors at once), and “Fill whole selection” in the Bucket tool. Hat tip to this dude and his commenters.
One last step: I thought the bare, unfilled yellow version looked too flat, so I tried different levels of fill to make the skull pop out from the background. I didn’t use bucket fill here–too fiddly with so many dots and edges. Instead I created a new layer of solid yellow and dropped the opacity to 17%. Then went to the drawing layer and used the magic wand tool to select the whole non-skull background. Then popped back to the yellow layer and cleared that selection, leaving yellow fill only in the boundaries of the skull outline. I also tried 10% and 25% opacity for the fill layer, but 10% was too subtle and 25% was starting to swamp some of the detail in the drawing. Between goofing around with colors and opacity levels we went through 10 versions at this stage, of which the one above is the ultimate champion.
So, that’s how the cover art came to be. Back to the book. There’s a bibliography with 1237 references (Vicki knows), and an index. The book is hardbound, with a printed cover and no dustjacket, and IMHO reasonably priced at $65, currently a few bucks less on Amazon. You probably already know whether you want a copy. If so, do the right thing–it’s not too late to get it by Christmas.
April 25, 2013
Generally when we present specimen photos in papers, we cut out the backgrounds so that only the bone is visible — as in this photo of dorsal vertebrae A and B of NHM R5937 “The Archbishop”, an as-yet indeterminate Tendaguru brachiosaur, in right lateral view:
But for some bones that can be rather misleading: they may be mounted in such a way that part of the bone is obscured by structure. For example — and this is a very minor case — the ventral margins of the centra in the photo above are probably slightly deeper than they appear, because the centra are slightly sunk within the plinth that holds the vertebrae upright.
So I’ve been toying with a different idea: instead of cutting the background out completely, leaving it in place but toning it down. Then the supporting structure is visible, but clearly distinct from the actual bone. (For a more extreme case, see the “Apatosaurus” minimus sacrum.)
Here’s how the image above looks if I desaturate the background:
I’m not sure what to make of this. It looks a bit strange to me, but that might only be the strangeness of unfamiliarity.
And it might not work so well (or indeed it might work better) for photos taken against a busier background.
What do you think?
April 18, 2012
I’ve been thinking about Barosaurus lately.
<homer>Mmmm … Barosaurus …</homer>
The best (and only, really) good recent treatment of Barosaurus is in John McIntosh’s chapter of the 2005 IUP Thunder Lizards volume. The main weakness of that chapter is that, while a lot of material is illustrated, the figures are rather small and not particularly well reproduced — and, in the case of the two-page spread of dorsal vertebrae, monumentally confusing:
Quick! find the fourth dorsal in posterior view! [˙ʇɟǝן ɯoʇʇoq ʇɐ s,ʇı :ɹǝʍsuɐ]
To get a better sense of the variation along the column, I scanned the two pages, loaded them into GIMP, joined them together, cleaned up the “white” of the background while retaining all the contrast I could, then moved each of the 27 illustrations to its own layer. Then I was able to rearrange them to my liking, align them, and produce this modified version:
So here it is for anyone else who finds it useful.
(One character that varies sequentially is of course the degree of neural spine bifurcation. But we won’t be flogging that dead horse any more — we’re done blogging, and the paper is in prep.)
- McIntosh, J.S. 2005. The genus Barosaurus Marsh (Sauropoda, Diplodocidae); pp. 38-77 in Virginia Tidwell and Ken Carpenter (eds.), Thunder Lizards: the Sauropodomorph Dinosaurs. Indiana University Press, Bloomington, Indiana, 495 pp.
March 3, 2012
In among all the open-access discussion and ostrich-herding, we at SV-POW! Towers do still try to get some actual science done. As we all know all too well, the unit of scientific communication is the published paper, and getting a submission ready involves a lot more than just the research itself. One of the most important aspects is preparing the illustrations — indeed Matt once told me that he thinks one of the best ways to put a paper together is to start with the illustrations, then write the text around them.
[Illustrations are often referred to as "figures". I don't know how the tradition got started, but since that term also means numbers, I will try to avoid it. If I tell you "I am working on the figures for my diversity paper", you don't know if I am accumulating statistics or preparing illustrations.]
Done well, illustrations can be things of beauty as well as scientifically informative.
There are a few things to be said about preparing good illustrations, so we’re kicking off a short series on the subject. This is the first.
But the zeroth was published here a couple of years ago. Since the most important illustrations in many palaeontology papers are those of the specimens, the base you’re working from is your specimen photographs. So you might want to refresh your memory by reading Tutorial 8: how to photograph big bones before we proceed.
There are various steps in getting from a photo to a finished, publishable figure, and we’ll look at those along the way. But somewhere along the line, if you’re publishing in a conventional journal such as the Journal of Vertebrate Paleontology, you’re going to flatten your colour images down to greyscale. Postpone that step till the last possible moment.
That should be too obvious to need saying, but I’ve got it wrong myself. When I was preparing the specimen photographs for the Xenoposeidon paper, destined for Palaeontology, I flattened the images too early in the process, with the result that the greyscale versions of the figures that were included in the paper are the only versions in existence. The upshot is that if you look at the full-resolution illustrations in the unofficial supplementary information, you’ll see that the version of Figure 3 available there is greyscale, just like the one in the paper.
By the time the three of us did our neck-posture paper in Acta Palaeontologica Polonica, we weren’t quite so dumb. So although the illustrations in the published paper are all greyscale, the two that are based on specimen photographs, rather than assembled from previously published greyscale components, were prepared in full colour, then flattened as the very last process before submission. As a result, the full-resolution illustrations in the unofficial supplementary information have figures 1 and 2 in colour:
So we were pretty happy with that. But by the time we came to submit the Brontomerus description a couple of years later, we’d had a rather obvious (in retrospect) thought: just because we can’t have colour in the printed journal, does that mean we can’t have it in the PDF? We asked the good people at Acta Pal. Pol., and they agreed that we could submit colour illustrations, they’d use them in the PDF, and then flatten them to greyscale themselves for the printed edition.
Since about fifty times as many people see the PDF as see the printed journal [yes, I just made than number up out of my head], that solution suited us very well. The outcome was the the PDF has gorgeous figures like this one:
(I’m slightly sorry to be displaying all our own illustrations here, but they do make the point and frankly I like looking at them. Especially that beautiful caudal vertebra.)
Why am I making such a big deal about colour? Because colour is information, and as scientists we love information. When you flatten a colour image to greyscale, you lose information, and that should never be done without regret. It’s perfectly possible that adjacent regions of a fossil will be a different hue but the same brightness: flatten the image and the two colours look the same, but in the original you can see a distinction. That’s valuable.
So in this day and age, The Right Thing is:
- Prepare your figures in colour
- Submit them in colour
- If the journal has a printed edition (and charges extra for colour printing, as most do), tell them to flatten to greyscale.
On the other hand, if you’re submitting to an open-access journal — and you should be, if you want to be widely read — there’s a good chance that it’s online-only (as with PLoS ONE and Palaeontologia Electronica), in which case the use of colour is a complete non-issue. The only reason to prepare monochrome figures then is (as with the Taylor et al. 2011b sauropod-neck bestiary above) when you’re constructing them from pre-existing greyscale images.
Taylor, Michael P., Mathew J. Wedel and Richard L. Cifelli. 2011a. A new sauropod dinosaur from the Lower Cretaceous Cedar Mountain Formation, Utah, USA. Acta Palaeontologica Polonica 56(1):75-98. doi: 10.4202/app.2010.0073
Taylor, Michael P., David W. E. Hone, Mathew J. Wedel and Darren Naish. 2011b. The long necks of sauropods did not evolve primarily through sexual selection. Journal of Zoology 285:150-161. doi: 10.1111/j.1469-7998.2011.00824.x
If you’re a scientist, then one of the things you need to do is prepare high-quality images for your papers. And, especially if you’re a palaeontologist, or in some other science that involves specimens, that’s often going to mean manipulating photographs. So image editing has become one of those “grey skills”, like word processing and phylogenetic analysis, that you need to have a little of, even if you’re not specialising in that direction.
Here at SV-POW!, none of us is anything remotely approaching wizardly when it comes to image-editing. But we’ve done enough of it that we have a few tips to pass on, so this is the first in an occasional series that will offer some random but relevant hints. (Matt and I both use GIMP, a free image-editing program, but I’m sure PhotoShop has the all the same facilities and more.)
Today: thirty-second colour-balancing. It’s a technique that comes in handy every now and then, especially if you take a lot of specimen photographs in poorly lit basements that make everything look greenish. It came up because in the previous post Matt included this photo of a partially dissected turkey neck:
All the orange made my eyeballs hurt.
So you can spend hours on colour-balancing a photo carefully, and that can be appropriate if you’re preparing a figure for publication. But to fix a photo like this one in thirty seconds, here’s what I do.
Load the image.
Bring up the Layers window and use it to duplicate the layer:
With the top layer selected, choose Colours -> Auto -> Equalize. (There is also a Colours -> Auto -> White Balance option, but I never find that it gives good results.)
Equalize will make the top layer look truly horrible:
Now go back to the Layers window, and play with the top layer’s opacity, so that you get a blend of the original and equalised images:
In this case, I found that 50% opacity looked about the best:
(While it’s still no oil-painting, it’s much better than the all-orange-all-the-time original.)
With the top layer still selected, choose Layers -> Merge Down to make the layers into one, and save the result.
It really does take about thirty seconds total, including the time to start up and shut down the image editor. (Yes, GIMP starts up more quickly than PhotoShop!)
Update (11 April 2012)
If you’re wondering why this is “part 0″, it’s because it was originally posted as a stand-alone article, and we only realised much later that it fits into the tutorial sequence — in particular, the planned multi-part tutorial on preparing illustrations.