Tilt-Shift Photography and Verified Views

Posted by in Photography on 20th January 2020

Tilt-Shift Camera

Why would you need a lens that tilts and shifts? What effect does it make? And why can’t you just use a normal lens when photographing buildings?

Well to start with, the name ‘tilt-shift’ is somewhat misleading in this case. It refers to a lens that is capable of both tilting and shifting, independent of the rest of the camera. The tilting motion is useful for miniaturising, a process that makes photos of landscapes appear to be tiny model versions, but it is a feature that is completely ignored in the realm of architectural photography.

Instead the ‘shift’ effect is useful but to understand why, it is important to lay some groundwork.

Think back to art lessons at school. When you first learned to draw things that were ‘3D’, you were likely taught about vanishing points. In other words, you draw two straight lines that are angled slightly towards each other and they will eventually converge. This is how we produce the 3D effect in two dimensions.

The problem is that, when you take a photograph, often the camera will create this effect when you do not want it. When photographing buildings or landscape scenes, we want all of the vertical lines to be parallel to one another. We do not want the buildings to look like triangles sloping away from us, we want them to look like rectangles with clear parallel lines.

Vertical Convergence
Vertical convergence of lines and resulting ‘keystone effect’ when photographing a building while looking up.

The ‘key-stone’ effect

This is called the ‘keystone effect’. Because we are producing very technically detailed images, we need to accurately convey the visual impact of certain proposals. So, our aim is to capture what the eye sees as accurately as possible which means removing the keystone effect.

Sometimes, this effect is not an issue. If the camera is completely horizontal (not angled upwards or downwards at all) then the vertical lines won’t converge since the camera lens is not distorting them in any way. The problem is that often we aren’t taking photos of buildings that are low down on the horizon. Sometimes we need to take photos of very tall buildings.

If we were to just angle the camera upwards to take a picture, then the keystone effect comes back. Even the slightest convergence of the lines can throw off the accuracy of the photomontages we produce. So somehow, we need to take a photo of the site that looks up, without moving the camera itself. That sounds complicated but it is actually very simple.

Tilt-Shift Lens 24mm Nikon
Nikon 24mm Tilt-Shift Lens. One of the many lenses used by rbmp.

Using a tilt-shift lens

When photographing buildings, you use the ‘shift’ part of the lens. Imagine you take a photo of a building that is directly in front of you. You take good care to not angle the camera up or down at all but keep it perfectly level on the tripod. You take the photo but realise that the top of the building is cut off.

So instead of angling the lens upwards for the next photo, you just raise the camera directly upwards about 5cm or so, keeping it level the whole time. This is what a tilt-shift lens does. It just raises the lens up a few centimetres without compromising your camera’s carefully calculated position. Now you can take another photo again with perfect parallel vertical lines.

Now across these two photos you can see the entire building. The first photo shows you all of the bottom half but misses the top, the second photo shows you the top half and misses the bottom. Using photo editing software, it’s possible to stitch these two photos together. Because the camera (and sensor) itself has not moved, the perspective across the two of them is totally uniform so there is little correction to be done to make them match up perfectly.

Because we use these photographs to create photomontages there are additional steps that get added into the process. We use these photos along with 3D models to create a visual representation of what proposed developments will look like. This means that once we have our shifted and stitched photo, we have to make sure that our 3D model lines up correctly with the perspective.

Tilt-Shift Photographs Combined
The two photograph: one level and two shifted up are combined to give photograph without any ‘keystoning effect.’

Use of tilt-shift photographs in verified views

In a similar way to how we work with panoramic images and stitching, with one perspective in these photos but two. So when we are setting up where the cameras are in our modelling software, we need to set up two cameras for each exact position that the lens was in each photograph in life. One at the original height, the other a few centimetres above it.

We then will render both images of the 3D model and do the same process of stitching those two perspectives together and layer them into the real-world photograph(s). It is important therefore to be as precise and as meticulous as possible when taking note of camera conditions on site as any small alteration or inconsistency can require a fair bit of post-production to fix.

This is how we work in certain locations. It is a level of detail that could be completely ignored in most amateur photography and absolutely vital when following the detailed methodologies set out by the Landscape Institute.

Tilt-Shift Photographs Combined with Model
The two photographs combined to give the full extent of the building within the Verified View.

Tilt-shift lenses are technical and expensive and serve a very specific function in architectural photography. There is software that can copy the effect for a fraction of the price for example automatically correcting vertical distortion. In most cases, a bit of distortion on tall buildings is hardly noticeable or perhaps even desirable for certain artistic effects.

But we are not trying to create an artistic effect in our photography. We are trying to create photorealistic technical images that are as accurate as possible to real life measurements. Therefore, it is paramount that we do capture the scene as precisely as possible, eliminating any camera distortion (that goes beyond the natural distortion of the human eye) and not relying on software to fix things as it can never do so perfectly.