In the first part of the photogrammetry tutorial I listed all the equipment you need. Now, it is time to look at the basics of photography as applied to the method: what are the general set-ups, and how do you have to adjust your camera settings to take optimal photos. And how do you judge the crop.
Depth of field
Whatever you want to make a 3D model of, the first thing you need to understand is how to produce photos with a suitable depth of field. Basically, this means getting as much of the object into proper focus, instead of having the parts closest to and farthest from you all blurry, and only the middle ground properly in focus. Or, if you use a point&shoot camera, whatever happens to be near the point of focus (i.e., often in the centre of the image) perfectly crisp, and all the rest just a washout.
Depth of field is influenced by several factors, for practical purposes (i.e., thing you can do something about) these are
And that’s it! Yes, indeed – that’s the one thing you can do directly on your camera to get a huge depth of field!
Now, there are a bunch of things you can do aside from that, but let’s stick to the basics for now. And there are a bunch of things you can adjust, but that you should not touch. More on that later.
F-Number, what does that do? It means making the aperture of your lens smaller or bigger. The bigger, the more light per time gets into the camera, and the lower will your depth of field be. Thus, you want a very tiny aperture, and that means very little light gets in. This increases the time the aperture must be open, i.e. the time it takes to take a well-exposed photo. Hence the need for a tripod: unless you have very unusual lighting conditions (basically, the sun directly behind you), you normally won’t be able to hand-hold your camera. Exceptions are flat objects that do not need a massive depth of field, like track fossils.
One thing to keep in mind is that you can change the focal length you use – which directly changes the distance to the specimen that you have to keep. Focal length also directly influences how much light gets into the camera. The longer, the less light – thus it can be a good idea to step closer to the object and shorten focal length. If you use a tripod, that doesn’t matter much, though: there is no big difference between a 5 s and a 15 s exposure time. And stepping close and using a wide view (18 to ~35 mm) can carry it’s own risks (photo alignment will sometimes fail), so personally, I rather go for something between 35 and 80 mm and live with the longer exposure time.
So, Rule #1 is: Maximize your f-number! It gives your greater depth of field, thus better images for photogrammetry.
Rule #2 stem directly from Rule #1: Use a tripod!
On to the other things you can do:
Obviously, for the same f-number your exposure time will be shorter if there is generally more light. So don’t try photogrammetry in a dank basement without turning on the room lights. Or bring additional light sources, if you can set them up. You can also, if you really really REALLY have to, up your ISO. Don’t ever think of going up from base ISO (typically 100) unless you really must. But yes, higher ISO numbers mean shorter exposure time, thus if you can’t use a tripod using ISOs up to 800 may save your day.
Also, obviously, you can just buy a faster lens. Be careful, though – you’ll easily spend several hundreds or more of bucks to gain a reduction in exposure time from 25 s to 5 s. Still way out of the window of hand-held shooting, thus irrelevant. Remember Rule #2!
Besides being out-of-focus, the easiest way for a point on your specimen not to show up in the model is to have it under- or overexposed. If your photo has washed-out highlights, the program will not find points in these areas, nor will it pick up blacked-out shadows. So you have to try to get all parts of the specimen well lit, and that without overexposing other parts.
Again, long exposure helps, and thus high f-numbers. Simply because it allows you to shoot without strong lights on the specimen, which invariably will create strong shadows. Especially if specimen are covered with lacquer, or otherwise have very smooth surfaces, on which directional light creates reflections.
So how do you find the correct exposure? The first step is using your camera’s automatic suggestion as a starting point. Do not auto-expose, but use the manual mode (M program on Canon cameras). Tap the shutter release and the camera will suggest an exposure time for the aperture and ISO settings you chose. Now check out your specimen and its surroundings. If you are using a turntable and a white background, and the specimen doesn’t fill your picture very well (e.g. because it is long and narrow), overexpose a bit, especially for dark specimens. If you have a darker background and a light-coloured specimen, underexpose a bit. Take a test shot and check it out on the computer screen – you’ll quickly get a feel for what’s too light and what’s too dark.
Thus, Rule #3: aim for a balanced exposure! This can sometimes involve moving your camera that tiny bit that hides a spotlight in a museum behind a pillar or even part of the skeleton you are trying to photograph. Or timing your photo session so the sun is on the other side of the building, and therefore doesn’t cast light and shadow patches on the mount.
Obviously, sometimes it will be very difficult to find the perfect exposure. Dark bone in cream-coloured limestone would be one of those things where you’ll likely end up with problems whatever you do. Here, it is possible to take two shots, one slightly over-, the other slightly under-exposed. Later, you can mask the bad parts of each image. But depending on what camera you buy there may even be a more elegant solution.
HDR-photos for photogrammetry
HDR captures a greater dynamic range between the lightest and darkest areas of an image than normal photography. HDR photos therefore can resemble the impression you get a lot more than any single photograph can. Basically, if you have a bright sky and a dark ground, you will see more of each than in any single photo, so you need to combine at least two photos to create the same effect. That’s what HDR programs do. So why not use such photos?
The big problem is that most HDR programs (or HDR functions in photo editing programs) do not spit out photos with “correct” EXIF information. But you need EXIF information for photogrammetry. Also, creating HDR images from several shots takes quite some time. Also, it can be quite an iffy task of fiddling with lots of parameters to achieve realistically coloured images – remember that all images you use need to “look” the same with regards to hue and tone, or photogrammetry will not work because no matching points can be found. Thus, the classic way of taking over- and under-exposed shots and calculating one image from them is something you should only do if you really have to. Luckily, some cameras have a built-in HDR mode, which creates images from usually three shots. If you use a tripod (here we go again!), the motion between the shots will be minimal, thus there will be no large crop and practically no loss of sharpness, and the photos can be used for photogrammetry.
So why did I not say right from the start that HDR photos are the way to go? Because they aren’t! For one thing, most automatic HDR modes do not vary exposure time, but ISO. Which means that one of the usually three photos that go into the final image is shot at a high ISO value, with all the dire consequences. Lotsa noise, which does show in the final photo. If you are rich and buy an really expensive DSRL, however, there often is a manual mode in the recent models that allows varying exposure time, and you will end up with perfect pictures.
Also, the blended images tend to be somewhat less crips, which can throw your point-detection algorithm. Not good, it will ruin your model. Overall, you’ll just have to experiment with your camera.
Thus, Rule #4: Try unconventional stuff, including HDR, if things don’t work well otherwise!
One thing that many people miss out on is that there is no need at all to hold your camera perfectly horizontal or vertical. Turn it any way you like, for photogrammetry it doesn’t matter! Thus, if you specimen is longer than wide, turn the camera so that the long axis fills the diagonal of your crop, and gain some resolution! The bigger the specimen in your photos, the higher will the maximum possible resolution of your model turn out. Thus,
Rule #5: maximize the size of your specimen in your images!
An expansion of this rule means that you should take a handful of overview images of a very large specimen, then get lots of closer-up shots of all parts, in order to get high resolution of the actual surface(s). I’ll explain that in detail in another post.
OK, next step is explaining the setups you can use for different types of specimens, and how to take pictures of actual specimens.