TPS Camera Recovery Procedure (TPS-CRP)

A step-by-step guide to recovering a stop-motion camera setup using the Three-Point System

Version 0.1 — Experimental Procedure

Introduction

You’ve carefully composed your stop-motion shot. Your camera is positioned, your lighting is established, and your set is ready.

Then something happens.

Perhaps the camera gets bumped. Maybe you need to remove it for maintenance, change equipment, or return to a shot after several weeks.

Now you need to reproduce the original camera view.

The TPS Camera Recovery Procedure (TPS-CRP) provides a structured method for approaching that problem.

It uses the Three-Point System (TPS), recorded camera measurements, and a reference overlay to guide the physical alignment of a camera with its original view.

Instead of making arbitrary adjustments until the image looks close, TPS-CRP organizes the recovery process into a sequence of observable, measurable steps.

The objective is to reproduce the original camera view as closely as practical—not to claim that a unique mathematical camera pose has been recovered.

TPS-CRP v0.1 is currently an experimental procedure. Its complete recovery sequence has not yet been physically validated as a repeatable Dragonframe workflow.

1. Understanding Camera Recovery

A camera’s view depends on several interacting variables.

Camera translation

The camera can move along three spatial axes:

1. Horizontally
2. Vertically
3. Forward or backward

Camera rotation

The camera can also rotate around three axes:

1. Pan: Horizontal rotation
2. Tilt: Vertical rotation
3. Roll: Rotation around the camera’s viewing axis

Together, these represent six degrees of freedom.

Optical variables

Other settings affect the image:

1. Focal length
2. Focus distance
3. Sensor dimensions
4. Active sensor area or crop
5. Lens distortion
6. Focus breathing

Changing one variable can produce an image difference similar to changing another.

For example, a camera moved farther from the subject may produce a smaller image. Reducing focal length can produce a similar change in apparent scale.

This is why recovering a camera cannot reliably be reduced to correcting a single visible discrepancy.

TPS-CRP approaches the problem by restoring recorded conditions and comparing the resulting image against a known reference.

2. The Sacred Triangle

The foundation of TPS is a set of three stable physical reference points.

These points are identified as:

A — First reference point

B — Second reference point

C — Third reference point

Together, they form the physical reference triangle, sometimes called the Sacred Triangle.

The distances between these points are recorded as AB, AC, and BC.

The corresponding points in the reference image are identified as:

a — Image projection of A

b — Image projection of B

c — Image projection of C

This distinction is important.

The physical triangle exists in three-dimensional space. Its image is a two-dimensional perspective projection.

The projected image triangle does not necessarily have the same shape as the physical triangle.

TPS preserves the correspondence between the physical points and their image projections.

During recovery, we compare the original projected points with their positions in the current camera image.

3. The A-Side and B-Side

TPS camera recovery involves two camera views.

A-Side: The Original Reference

The A-Side is the recorded camera setup you want to reproduce.

It includes the original reference image and the measurements and camera information used to document that setup.

The A-Side establishes the target.

B-Side: The Recovery Setup

The B-Side is the current camera view.

During recovery, the camera is physically adjusted to bring the B-Side image into agreement with the A-Side.

The objective is to align the corresponding reference points while preserving the correct image registration.

The Fundamental Comparison

At any moment during recovery, we can examine the differences between:

1. Original point a and current point a
2. Original point b and current point b
3. Original point c and current point c

These differences provide feedback about the remaining image mismatch.

However, they do not independently identify a unique physical correction.

Several camera adjustments may affect the same projected point.

4. Equipment and Information Required

Before beginning, gather the following.

Reference Materials

1. Original A-Side reference image
2. Saved TPS data
3. Transparent reference overlay
4. Recorded physical measurements
5. Original image dimensions
6. Identification of reference points A, B, and C

Camera Information

1. Camera manufacturer and model
2. Lens information
3. Recorded focal-length information
4. Focus setting or distance
5. Camera tilt
6. Camera roll, when available
7. Sensor and capture-mode information

Physical Measurements

1. A–B Base Pair measurement
2. Other recorded triangle measurements
3. TPS Reference Distance
4. Relevant camera support or stage measurements

Camera Alignment Equipment

1. Camera and appropriate support
2. Adjustable camera positioning equipment
3. Measurement tools
4. A live-view or image-comparison environment

The experimental TPS-CRP workflow uses Dragonframe for reference-overlay comparison.

Other environments may support similar comparisons, but their implementation details must be verified independently.

5. Preparing the Camera

Before attempting fine alignment, restore as much of the original camera configuration as possible.

Use the recorded camera and lens settings.

Check that the camera is operating with the intended image dimensions, orientation, and capture mode.

Restore the focus setting and approximate camera position.

Make sure the physical reference points are still identifiable and have not moved relative to one another.

Why This Matters

If the original image was captured using one sensor crop and the recovery image uses another, the reference triangle may appear at a different scale even when the camera is physically close to its original position.

Likewise, changing focus can alter image magnification on some lenses.

These differences can be mistaken for positioning errors.

Before correcting camera geometry, verify that the image comparison itself is trustworthy.

6. Registering the Reference Overlay

The first operational task is to establish a reliable comparison between the A-Side and B-Side.

The TPS reference overlay represents the original image positions of points A, B, and C.

Step 1 — Open the Live Camera View

Connect the camera to Dragonframe and establish the intended live camera image.

Confirm the image orientation and capture settings.

Step 2 — Load the Reference

Load the transparent TPS overlay or appropriate reference plate into Dragonframe’s overlay or reference-image functionality.

The exact controls may vary by Dragonframe version.

Step 3 — Verify Image Registration

Confirm that the overlay and live image use compatible:

1. Image dimensions
2. Aspect ratios
3. Orientation
4. Crop
5. Scale
6. Coordinate systems

Do not stretch or resize the reference simply to conceal an unexplained mismatch.

Step 4 — Verify Point Correspondence

Confirm that each reference marker represents the correct physical point.

A must correspond to A.

B must correspond to B.

C must correspond to C.

An incorrect point correspondence invalidates the comparison.

Step 5 — Adjust Overlay Visibility

Choose an opacity or display mode that allows the original markers and live reference points to remain visible.

You should now be able to observe the displacement between the original and current projections.

Important: The overlay is a visual representation of the TPS reference, not a replacement for the recorded measurements and camera information.

7. The TPS-CRP Recovery Sequence

The experimental procedure follows eight stages:

1. Match tilt.
2. Match roll, when available.
3. Match distance using the A–B Base Pair.
4. Match horizontal position.
5. Match camera height.
6. Use C for pan and remaining alignment.
7. Fine-check all three points.
8. Obtain operator approval.

This sequence is proposed operational guidance, not a universal requirement of the Three-Point System.

Camera adjustments interact with one another. You may need to revisit earlier stages as the alignment improves.

8. Stage One — Match Camera Tilt

Begin by restoring the recorded camera tilt.

Tilt describes the camera’s vertical rotational orientation.

Use the original measurement and the same measurement convention wherever possible.

Procedure

1. Locate the recorded A-Side tilt value.
2. Identify how that measurement was originally taken.
3. Adjust the camera to reproduce the recorded tilt.
4. Observe the resulting projected triangle.
5. Record any uncertainty in the measurement.

Why Begin with Tilt?

Camera tilt affects the vertical projection of the reference triangle.

An incorrect tilt can produce image discrepancies that resemble translation errors.

Restoring tilt provides a useful starting condition for the remaining adjustments.

However, matching the recorded tilt does not prove that the complete camera orientation has been recovered.

Do not attempt to compensate for a known tilt error by moving the camera vertically.

Restore the documented orientation first, then evaluate the remaining image mismatch.

9. Stage Two — Match Camera Roll

Camera roll describes rotation around the viewing axis.

If the original roll is known, restore it using the recorded convention.

Procedure

1. Locate the original roll measurement.
2. Confirm the measurement’s sign and reference convention.
3. Adjust the camera roll.
4. Compare the orientation of the A–B Base Pair.
5. Check the position of C.

What to Observe

Roll can change the orientation of the projected triangle.

However, changes in pan, tilt, camera position, and perspective can also affect the apparent orientation of the reference points.

Therefore, the projected A–B line should not be treated as an independent roll measurement.

If the original roll is unknown, document that uncertainty.

Do not invent a roll value from the triangle alone.

10. Stage Three — Match Distance Using the A–B Base Pair

This is one of the central stages of the procedure.

The A–B Base Pair provides the initial scale and image-registration reference.

During recovery, compare the apparent separation of A and B in the current image against their separation in the original reference.

What the Base Pair Tells Us

The projected A–B pair provides several useful observations:

Separation: How far apart A and B appear.

Midpoint: The image location halfway between A and B.

Orientation: The direction of the line connecting A and B.

Common displacement: Whether both points have moved in a similar direction.

Together, these observations help organize camera alignment.

Procedure

1. Restore the recorded TPS Reference Distance as closely as practical.
2. Compare the current A–B separation with the A-Side reference.
3. Make a small distance adjustment.
4. Observe the effect on both points.
5. Repeat as necessary while monitoring the full triangle.

Interpreting Scale

If A and B appear closer together than in the original image, a distance difference may be involved.

If they appear farther apart, the opposite may be true.

However, this is not a unique diagnosis.

Apparent scale can also change because of:

1. Focal length
2. Focus breathing
3. Sensor crop
4. Image resizing
5. Lens distortion
6. Camera orientation

Do not assume every A–B scale mismatch is caused by camera distance.

The recorded measurements and optical configuration remain essential.

11. Stage Four — Match Horizontal Position

Once the camera’s orientation and approximate distance have been restored, begin refining its horizontal position.

The goal is to reduce the lateral displacement of the A–B Base Pair.

Procedure

1. Compare the current A and B positions with their reference markers.
2. Observe whether both points are displaced in a similar horizontal direction.
3. Make a small horizontal camera adjustment.
4. Observe how both projected points respond.
5. Continue until the Base Pair is usefully close to its target.

Why Small Adjustments?

Real camera supports do not always move along perfectly independent axes.

For example, moving a crane or rotating an arm may introduce both translation and rotation.

The visible result may differ from the nominal direction of the control.

Small adjustments allow you to observe the actual relationship between a camera control and the projected reference points.

This is particularly important when working with motion-control rigs or articulated camera supports.

12. Stage Five — Match Camera Height

Next, address the remaining vertical displacement of the reference points.

Camera height changes the viewpoint relative to the physical reference triangle.

Procedure

1. Compare the vertical positions of A and B.
2. Make a small camera-height adjustment.
3. Observe the effect on both points.
4. Check whether C improves or becomes more displaced.
5. Continue refining the camera position.

Height Versus Tilt

Camera height and tilt can produce related visual effects.

A vertical discrepancy does not automatically mean the camera is too high or too low.

This is why the procedure begins by restoring the recorded tilt.

If adjusting height improves A and B but substantially worsens C, reassess the overall configuration rather than continuing to move in the same direction.

13. Stage Six — Use Point C to Complete the Alignment

Once A and B are reasonably close to their reference positions, bring C into the comparison.

This is where the third point becomes especially valuable.

Two points establish the Base Pair, but they do not fully describe the projected reference triangle.

C supplies an additional image observation.

What C Reveals

Point C may expose discrepancies involving:

1. Camera pan
2. Remaining vertical alignment
3. Depth-related changes
4. Perspective differences
5. Coupled camera adjustments

Procedure

1. Confirm that A and B are usefully close to their reference positions.
2. Compare C with its original image marker.
3. Make a small pan or other justified camera adjustment.
4. Observe the effect on all three points.
5. Continue refining the complete triangle.

The Important Rule

Never evaluate C in isolation.

An adjustment that improves C may move A and B away from their targets.

The goal is not to align one point at the expense of the others.

It is to improve the correspondence of the complete projected triangle.

Does C Identify the Correct Camera Adjustment?

Not uniquely.

A displacement at C does not prove that pan, height, or distance is the sole source of the error.

The operator must consider the recorded setup, the behavior of the camera support, and the effects observed during adjustment.

14. Stage Seven — Fine-Check A, B, and C

At this stage, the camera should be approaching the original reference view.

Now evaluate the complete image.

Check the Three Reference Points

Compare:

1. A-Side a against B-Side a
2. A-Side b against B-Side b
3. A-Side c against B-Side c

Look for remaining differences in position and overall triangle appearance.

Check the Frame

Confirm that the image has not been:

1. Cropped unexpectedly
2. Rescaled
3. Mirrored
4. Rotated
5. Registered incorrectly

Check the Optics

Verify that the focal length, focus state, and capture configuration remain consistent with the intended recovery.

Check the Rest of the Scene

This is essential.

Even when all three reference points appear aligned, other stable features may not agree.

Inspect set edges, background details, and other stationary features visible in both images.

A successful three-point match is useful evidence, but it does not guarantee that the entire image is identical.

Continue Iterating

Make small adjustments while observing the effect on all three points and the rest of the image.

Avoid large corrections that destroy previously established alignment.

15. Stage Eight — Operator Approval

The final stage is a production decision.

The responsible operator evaluates the recovered view and determines whether it is sufficiently close for the intended shot.

There is no universal numerical tolerance established by TPS-CRP v0.1.

A satisfactory match for one production may not be acceptable for another.

Before Approving

Confirm that:

1. The reference points have been correctly identified.
2. The overlay is properly registered.
3. The camera settings have been reviewed.
4. All three projected points have been evaluated.
5. Other stable scene features have been inspected.
6. Remaining differences are understood or documented.

Record the Result

Preserve the recovered image and note any remaining discrepancies.

Operator approval means the result is accepted for the intended use.

It does not establish that the original six-degree-of-freedom camera pose has been uniquely recovered.

16. Troubleshooting Common Recovery Problems

A and B Align, but C Does Not

This indicates that the Base Pair comparison alone is insufficient to reproduce the full projected triangle.

Possible contributing factors include:

1. Camera orientation
2. Camera position
3. Optical differences
4. Incorrect point correspondence
5. Physical reference changes

Recheck the recorded setup and use small adjustments while monitoring all three points.

Do not automatically interpret C’s displacement as a single-axis error.

The Triangle Is the Correct Shape but the Wrong Size

Check the camera distance and optical configuration.

Verify focal length, focus state, active sensor area, and image registration.

A similar-looking projected triangle does not establish that the camera is correctly positioned.

The Overlay Will Not Line Up

Verify the original and current image dimensions.

Check for unexpected cropping, rotation, mirroring, scaling, or aspect-ratio changes.

Do not continue camera adjustments until the comparison frame is understood.

Moving the Camera Improves One Point but Worsens Another

This is expected during some stages of alignment because camera controls are coupled.

Reduce the adjustment size.

Revisit the recorded orientation and distance.

Evaluate the complete triangle rather than optimizing one point independently.

All Three Points Match, but the Scene Still Looks Different

Inspect the remaining image for changes in perspective, lens behavior, set geometry, or registration.

Three corresponding points do not establish complete image equality.

Record the differences and determine whether the recovery is acceptable for the intended shot.

17. Recording a Recovery Attempt

TPS-CRP is also a documentation procedure.

Each recovery attempt should preserve enough information to understand what was done and evaluate the result later.

Reference Information

1. Original A-Side image
2. Recovered B-Side image
3. TPS package
4. Reference overlay
5. Physical measurements
6. Point identities and image correspondences

Camera Information

1. Camera model
2. Lens
3. Capture mode
4. Image dimensions
5. Focal-length information
6. Focus information
7. Tilt and roll measurements

Recovery Observations

1. Overlay registration method
2. Camera adjustments made
3. Order of adjustments
4. Observed changes at A, B, and C
5. Differences elsewhere in the scene
6. Unresolved uncertainties

Final Result

Record whether the operator accepted or rejected the recovered view.

Preserve the original and recovered images for direct comparison.

These images provide the primary practical evidence of the recovery’s success.

18. What TPS-CRP Does Not Claim

The Three-Point System is a practical geometric reference method.

TPS-CRP is an experimental procedure for using that reference during physical camera recovery.

It is important to distinguish its purpose from other camera-solving techniques.

TPS-CRP does not currently claim to:

1. Recover a mathematically unique camera pose.
2. Independently solve all six camera degrees of freedom.
3. Calibrate a lens.
4. Eliminate perspective ambiguity.
5. Guarantee exact image equality.
6. Guarantee that all scene features will align.
7. Achieve a universal numerical recovery tolerance.

The procedure provides a structured method for reducing visible differences between an original camera view and a recovery view.

Its effectiveness must be evaluated through documented physical testing.

19. Current Validation Status

TPS-CRP v0.1 remains experimental.

The proposed sequence has not yet been physically validated as a complete, repeatable Dragonframe camera-recovery procedure.

Existing LIMNTPS overlay testing demonstrates practical use of a reference overlay, but it does not establish the accuracy or repeatability of the entire TPS-CRP sequence.

Further validation will require controlled recovery attempts with preserved reference images, measurements, camera adjustments, and results.

Until that work is completed, the procedure should be understood as documented experimental guidance.

20. The Purpose of TPS-CRP

Camera recovery has traditionally depended heavily on the experience and judgment of the camera operator.

TPS-CRP does not attempt to eliminate that expertise.

Instead, it provides a framework for applying it systematically.

By preserving reference geometry, restoring measured conditions, and comparing observable image differences, the procedure makes the recovery process easier to document, communicate, evaluate, and improve.

The long-term goal is to make reliable camera recovery more accessible to stop-motion artists, educators, and production teams.

Three points. One reference. A repeatable process to investigate and refine.

Learn More

Related LIMNTPS Articles

1. Getting Started with LIMNTPS
2. Understanding the Three-Point System
3. Choosing Reference Points
4. Saving Your A-Side
5. Exporting TPS Files and Overlays
6. Recovering a Camera Position

Support:  ⁠LIMNMEDIA Support Portal

Community:  ⁠LIMNMEDIA Discord

Developed by LIMNMEDIA

Less guesswork. More animation.