Optics 034 · Imaging, instruments, and visual systems

Tilt-Shift & Scheimpflug Studio

How does tilting the lens rotate the plane of focus and reshape depth of field? Once the lens tilts, the sensor plane, the lens plane and the plane of sharp focus all meet in one line, the Scheimpflug line. The plane of focus tilts with it and can lie along a sloping subject such as a tabletop or a road. Refocusing then swings that plane around a second line, the hinge line, and the depth of field becomes a wedge instead of a slab.

Tilt a lens with a finite pupil to bring a fixed, sloped checkerboard into focus. In a ray-traced model city, tell lens shift apart from camera pitch, and create selective geometric blur without shrinking the buildings or adding a Gaussian filter.

Focus-plane angle / axial intercept

\psi,s_0

$0.50$

Construction lines / perspective ratio

y_S,y_H,r

$50\%$

Target-point blur / selected tolerance

b_{\max},c

$0.00\pi$

Model regime $\text{valid model regime}$

The model uses an ideal projective thin lens and geometric images integrated over the pupil. Purple surfaces are sampled boundaries of the blur threshold. Diffraction, aberrations, the image circles of commercial lenses, exposure and the accuracy of real measurements are not simulated.

Physics tutorial

Audit the physics of Tilt-Shift & Scheimpflug Studio

BackgroundTilting the lens changes which plane in the scene is imaged sharply on the fixed sensor. The distances in the thin-lens law are measured perpendicular to the lens, not along any horizontal line. With the sensor perpendicular to the camera axis, the angle of the focus plane satisfies $\tan\psi=\frac{v\sin\alpha}{v\cos\alpha-f}$ . Raising the lens does something else: it moves the field of view relative to the fixed sensor. Refocusing changes the distance between lens and sensor.

Why it mattersHow does tilting the lens rotate the plane of focus and reshape depth of field?

Start with the essentials

Focus question: How does tilting the lens rotate the plane of focus and reshape depth of field?
One-sentence intuition: The lens, sensor and focus planes meet on the Scheimpflug line. Refocusing pivots the focus plane around a different line, the hinge. Their signed heights in camera coordinates are $y_S=h_L-v\cot\alpha,\qquad y_H=h_L-\frac{f}{\sin\alpha}$ . These are two different lines. As the tilt goes to zero, both move off to infinity, and no finite marker is drawn for them.

Core mathematical model

Conjugate distances measured perpendicular to the lens

p=-\mathbf n\cdot(\mathbf P-\mathbf L),\quad q=\mathbf n\cdot(\mathbf S-\mathbf L),\quad\frac1p+\frac1q=\frac1f

The lens normal tilts with the lens. In this ideal thin-lens model, with the same index on both sides, the ray through the lens center keeps its direction. Every ray sampled across the pupil goes to the same conjugate point.

Focus plane in camera coordinates

(v\cos\alpha-f)x+v\sin\alpha(y-h_L)+vf=0

The sensor stays at a fixed distance along the axis and keeps a fixed orientation. Requiring the thin-lens image of each object point to land on the sensor gives this plane, with lens rise included.

Finite pupil and geometric defocus

a=\frac{f}{2N},\qquad b_{\mathrm{untilted}}=2a\left|1-\frac{v}{q}\right|,\qquad b_g\le c

The formula for an untilted lens on axis is an independent reference. With tilt, the spots are traced to the actual sensor plane. The readout takes the largest separation between paired rim rays at three marked ray origins. The purple boundaries mark the chosen blur tolerance; they do not represent diffraction or human visual acuity.

Converging verticals from pitch, compared with shift

r_{\mathrm{pitch}}=\frac{d\cos\beta+y_{\mathrm{low}}\sin\beta}{d\cos\beta+y_{\mathrm{high}}\sin\beta},\qquad r_{\mathrm{level}}=1

This ratio of chief-ray widths compares the top and the bottom of a fixed model facade. With the sensor level, vertical lines stay parallel and the shift only moves the crop. Shift does not digitally straighten a photo already taken with the camera pitched up, and it does not enlarge the image circle.

Common difficulties

Shift causes miniature blur and tilt corrects all perspective

Typical misconceptionTilt and shift are two names for the same effect, so a blur filter can reproduce the whole optical experiment.

Better mental modelTilt changes where the plane of sharp focus lies. Shift changes the framing. Pitching the camera changes the perspective. Here the blur grows with depth because rays are integrated over a finite pupil, and the model buildings keep their size. No Gaussian filter is applied afterward.

Run the experiment

01
Scene 1: Scheimpflug intersection
Start with the sloped checker. Set the lens tilt to zero, then return it to six degrees. Sweep the focus extension and watch the focus plane pivot about the fixed hinge while the target stays put.
What to observe: Both images share the same physical sensor field, the same linear radiance scale and the same viewing rotation. Neither is normalized or rescaled on its own.
02
Scene 2: Architectural shift
Open Architectural shift. Sweep the lens rise, then the pitch of the reference camera. Compare the change in framing with the converging verticals, and read each top-to-bottom width ratio.
What to observe: At zero tilt the focus plane is parallel to the sensor and the construction lines are at infinity. When the focus becomes singular, the readout says so instead of showing a huge finite focus distance.
03
Scene 3: Miniature-effect wedge
Open Miniature. Reverse the tilt and stop down: the sharp region moves and the geometric blur shrinks. Then raise the acceptable blur tolerance. The purple boundaries move, but the image stays the same.
What to observe: A higher f-number reduces geometric blur here. Real cameras also lose light and eventually become diffraction-limited; neither exposure nor diffraction is simulated.