Capturing trace images (or digital textures) with minimal perspective distortion


In 3D-modeling, good trace images and blueprints are a great way to ensure that your model will be accurate, and they can make the modeling part much easier. Blueprints found on the net aren't necessarily always accurate in all of their four views though (side,front,back,top), so capturing photographs of the real thing, when possible, is usually the most accurate option. However, when using photographs as trace images, perspective distortion can become a problem. For example, parts of a car modeled solely in the side view, might be slightly off when you compare those with the trace images of the front/back.

There are two keys for capturing great, flat, textures and trace images with minimal distortion. The first is Distance, and the second is Center Everything.


Distance

The further you are from the subject being photographed, the lesser the distortion from perspective will be. The more distance you can put between yourself and the vehicle (for example), the better: your images will become flatter, which is exactly what we're after for textures and trace images. However, this means that you will need plenty of space, so you'll want an appropriate location for the shoot, such as a mall parking lot, or something as flat and spacious. Make sure that photography is allowed at whatever location you choose.

Of course, you don't want the vehicle to cover only 2% of the frame. The camera you're using probably has some sort of an optical zoom, so that's a start. Additionally, you can use a feature called something like "smart digital zoom", which uses the available pixels on your camera to crop the view (although this also means that you're not getting the full resolution image your camera is capable of, so mind that). Note that the standard, "unsmart" digital zoom is different in that it simply enlarges the pixels, which is of no use in most situations, and should be avoided here as well.

If you have a DSLR available for you, such as a Canon 600D, you can go further. The cheaper models use a smaller sensor which is often called a "crop sensor" ("1.6x crop" is common in most Canon DSLRs), while the more expensive models use a bit larger sensor which is often called "full frame". The difference is that a larger sensor yields a wider view using a certain lens, so you actually want the cheaper model utilising a smaller sensor, for this task.

DSLRs also use interchangeable lenses, which means that you can swap the standard 18-55mm walkaround lens with something like a 70-300mm telephoto lens, which has considerably more "reach". The higher the focal length, the better it is for this task. If you own a teleconverter instead of a longer lens, you can use it to multiply the focal length of a lens you do own, by a fixed value (1.4x and 2x are most common). Of course, you can also use it on the long lenses, and some teleconverters can even be stacked.

As for "smart digital zoom": DSLRs usually do not have this feature, but the equivalent to this is simply shooting at the max resolution your camera is capable of, and stepping further from the vehicle so that it will cover smaller part of the frame. As you then crop the excess areas on your computer, you get an image that was shot from a greater distance, but which still "fills the frame". Again, you are of course reducing the resolution from the maximum ("using the megapixels to zoom"), but this shouldn't be a problem as long as the remaining image is sharp and detailed enough for the tracing.

My last tip regarding distance is actually a tripod, since handholding the camera isn't really an option at these distances. And the self-timer feature, for an even more stable capture.


Center Everything.

"Center Everything", as you might guess, is avoiding a camera position which forces you to shoot at a non-90 angle in relation to the vehicle side being photographed.

First of all, before you start walking around the parking lot, you want to make sure that the camera is placed at the midpoint of the total vehicle height. Place your tripod with the camera next to the vehicle, and adjust the height for the desired camera position. If the parking lot is dead flat, you want to lock this height and avoid any further adjustments to it, whether you're shooting the front, back, or side view.

With the height set correctly, all that is left is to make sure that the camera is positioned directly to the side of the vehicle when shooting the side view, directly in the front of it when shooting the front view, and directly behind it when shooting the rear view. If your camera has focus points in the optical viewfinder, or a grid overlay in the digital LCD, you can use them to aim the camera directly at the center of the vehicle as well. This way you'll avoid any pitch and yaw rotations. As for roll rotation: the more level the camera is, the better, but this can also be fixed in post-processing.


Post-processing

When you've captured the required images (including any textures requiring space to shoot), it's time to refine them a bit on the computer.

The images can probably use at least a very slight straightening, no matter how careful you were during shooting. If your image editor has a grid overlay, and an adjustable grid pitch, you can use these two helpful tools for this task. For the front and rear view, rotate the image until the symmetrical parts on both sides of the photographed vehicle, fall precisely on the same horizontal grid line. When the front and rear view images are properly straightened and the symmetrical parts are level, you should also be able to place the vertical centerline of the grid in a way that it runs along the middle of the vehicle. As for the side view, you can probably use the points where the wheels contact the ground, or any horizontally level feature in the bodywork (assuming you're 98% sure about it, or more), as a guide for the straightening.

You should now have trace images that are flat, leveled, and ready for use in modeling.


Random thoughts

- Overcast days are a great way to ensure that your textures (and photographs in general) will be neutral, and have a minimum amount of unwanted shadows, glare and highlights
- If you're just walking around with the camera and not using a tripod, consider a lens with an image stabilisation feature, which lets you capture images with visibly less blur from camera shake
- Visually accurate miniature models can also be useful for acquiring trace images as it's easier to setup the shoot, since the reduced size also means reduced distances for photographing them. However, as for the textures, it's practically impossible to shoot interior textures from those.
- Top view can be tricky to shoot with real vehicles, but fortunately it's not really required. However, if it just so happens that you know a high-rise building nearby, which has a parking spot right next to it, and allows access to a high vantage point directly above, then great! Lose that vertigo, put the camera neck strap on, and snap away. Otherwise though, you could try rolling the vehicle over so that it ends up on its side, hopefully in a way that it won't become too damaged, and that there's plenty of space around for shooting, all the while you're apologizing the other drivers for blocking the road... actually, on the other hand, perhaps it's not worth the trouble.