Optimizing PTGui Workflow for Mosaic 51 (6-Sensor 360° Rig) – Fixed Camera Rig Stitching Strategy?

47 views
Skip to first unread message

Cameron Rex

unread,
Sep 12, 2025, 12:01:01 PM (13 days ago) Sep 12
to PTGui Support

Hi all,

We're working on developing an efficient and repeatable workflow for stitching 360° panoramas captured by the Mosaic 51 camera using PTGui Pro, and are hoping to get some advice from the community or PTGui support directly. We are new to PTGui Pro (having peviously used MistikaVR), but are so far liking the features in PTGui Pro (that MistikaVR does not have). In our initial tests we have had pretty good success on over 2,500 panoramas.


Mosaic 51 Overview (Fixed-Rig, 6-Camera Panoramic System)

The Mosaic 51 is a professional-grade, vehicle-mounted panoramic camera designed for high-volume mobile mapping and asset documentation. It features:

  • Six fixed-position global shutter cameras, each with:

    • Sony IMX265 12.35 MP sensors (4054 × 3046 px)

    • 124° × 96° FOV with ~2.52 mm lens

  • Total stitched resolution: ~12.3K (12288 × 6144) for full equirectangular 2:1 output

  • Synchronization: Microsecond-level across all sensors (hardware sync)

  • Recording: Each sensor saves to its own high-bitrate H.264 .mp4 file at 120 Mbps

  • Triggering: Distance-based GNSS triggering available

  • Rugged design: Fixed geometry, sealed housing, vendor-unlocked format


Frame Extraction & Input Format

We extract frames from each of the 6 .mp4 files independently using ffmpeg, producing 6 synchronized JPEG images per frame. The result is a predictable, ordered set of images for each capture (e.g., _01.jpg to _06.jpg per frame), with no missing data or timing drift between views.

Each extracted frame set reflects a synchronized 360° capture.


Stitching Workflow in PTGui (Current Status)

Here's what we’ve implemented so far in PTGui Pro:

  1. Selected a reference frame and manually stitched it in PTGui

  2. Created a reusable .pts template based on this frame

  3. Applied that template to all frame sets using Batch Builder + Batch Stitcher

  4. Initially enabled “Align Images”, but now questioning whether this is appropriate, given the fixed nature of the rig

We also have a MistikaVR calibration file and are capable of exporting camera rig definitions as JSON, though it’s not clear how (or if) this data can be imported into PTGui.


Our Goals

  • Produce high-quality, consistent panoramic images at scale (10,000+ frames per project)

  • Minimize parallax/ghosting artifacts (especially with near-field assets)

  • Avoid unnecessary per-frame adjustments to maintain performance

  • Leverage existing rig calibration and minimize user intervention


Key Questions for the PTGui Team or Advanced Users

  1. Should we disable “Align Images” in Batch Stitcher?
    Given our fixed-rig design and a well-tuned template, is “Align Images” even necessary—or does it introduce risk of misalignment?

  2. What’s the best way to handle parallax?
    While our use case is mostly infrastructure (e.g., roads, buildings), we still get some ghosting on close objects (e.g., poles, signs). Are there PTGui-specific strategies (e.g., mask workflows, control point strategies) for dealing with parallax in multi-lens rigs?

  3. Can PTGui utilize external calibration data?
    We can export rig geometry from MistikaVR (FOV, offset, etc.) to JSON—can any of this be imported into PTGui or manually translated into lens settings and anchor points?

  4. How do we best lock down template parameters?
    Is there a preferred scripting/workflow approach to ensure that lens settings, yaw/pitch/roll offsets, and control points remain static across the entire batch?

  5. Can we globally apply masks across batch sets?
    We'd like to mask out the vehicle/roofline and apply that consistently. Can this be templated in a way that Batch Stitcher honors across all image sets?


Additional Notes

  • We’re using PTGui Pro, latest version

  • All 6 cameras fire simultaneously

  • Rig is fixed, weather-sealed, and consistent across capture sessions

  • No exposure drift (thanks to hardware sync + internal processing)


Thanks so much in advance for any guidance. Happy to provide example files or templates if useful!


Cameron

Message has been deleted

Cameron Rex

unread,
Sep 12, 2025, 12:21:44 PM (13 days ago) Sep 12
to PTGui Support
Here is the layout of the Mosaic 51 camera.

M51 visualization python script output

PTGui Support

unread,
Sep 12, 2025, 12:48:45 PM (13 days ago) Sep 12
to pt...@googlegroups.com
Hi Cameron,

I'm inclined to say disable Align Images. Create a template from a
project where you've not placed control points on any nearby objects, so
the template doesn't suffer from parallax.

You will have stitching errors on nearby objects. This is unavoidable
unfortunately. PTGui doesn't use optical flow or any other kind of
algorithm to reduce parallax errors, and there's not much you can do
about this. PTGui's optimum seam placement will help a bit hiding the
errors (if there's enough overlap).

The alternative would be to run Align Images. This will not fix parallax
either. In some cases the stitching errors in the foreground may be less
severe, but this is at the expense of introducing new stitching errors
in the background. The drawback is that you will get inconsistent
stitching (due to the differing effects of parallax between projects).
Also the Batch Stitcher will take longer, and there's always a risk of
PTGui not finding control points for some projects.

But that's just my guess. You could try running a batch both ways and
see how it goes.

> 3.
>
> *Can PTGui utilize external calibration data?*
> We can export rig geometry from MistikaVR (FOV, offset, etc.) to
> JSON—can any of this be imported into PTGui or manually translated
> into lens settings and anchor points?

No, you would need to do calibration within PTGui. This is very simple,
just give it a project with sufficient control points and it will
determine the best lens parameters. Again, for a bullet proof template I
think it's best if you make sure there are no control points on nearby
objects.

>
> 4.
>
> *How do we best lock down template parameters?*
> Is there a preferred scripting/workflow approach to ensure that lens
> settings, yaw/pitch/roll offsets, and control points remain static
> across the entire batch?

If you don't run Align Images, the template parameters are locked; each
project is stitched identically.

If you do run Align Images, you can configure the optimizer in the
Optimizer tab. Switch the Optimizer tab to Advanced mode, and check only
those parameters you want to be optimized. Save as your template.

> 5.
>
> *Can we globally apply masks across batch sets?*
> We'd like to mask out the vehicle/roofline and apply that
> consistently. Can this be templated in a way that Batch Stitcher
> honors across all image sets?

Yes you can define masks in your template. They will be applied during
stitching.

Hope this helps.

Kind regards,

Joost Nieuwenhuijse
www.ptgui.com

Erik Krause

unread,
Sep 12, 2025, 2:15:19 PM (13 days ago) Sep 12
to pt...@googlegroups.com
Am 12.09.2025 um 16:44 schrieb Cameron Rex:

> *What’s the best way to handle parallax?*

The best way to handle parallax is to minimize the lens distance. Some
manufacturers of multi-lens panorama rigs have apparently failed to
understand this.

> *How do we best lock down template parameters?*
> Is there a preferred scripting/workflow approach to ensure that lens
> settings, yaw/pitch/roll offsets, and control points remain static across
> the entire batch?
Control points are never copied because they are specific to a
particular image pair. Anything else can be set on Project Settings tab.
To really lock down parameters, avoid Align Images (as Joost wrote).

> *Can we globally apply masks across batch sets?*

Yes. On the Project Settings tab, under 'Template behavior', you can
specify what is copied when a template is applied. Pay attention to the
mouse pointer hints!

--
Erik Krause
http://www.erik-krause.de

Cameron Rex

unread,
Sep 15, 2025, 7:28:37 AM (11 days ago) Sep 15
to PTGui Support
Thank you. We have made good progress in fine tuning the process.
Reply all
Reply to author
Forward
0 new messages