1. Lightform Guide
  2. Guides
  3. Scanning

Disparity Tips

Avatar
Brittany Factura
Follow

Using visible structured light, we extract disparity information from your scene. Below we'll cover the expected results for 2D, 2.5D, and 3D scenes.

Disparity will work best on 3D scenes, but might not work well on scenes that are too flat. In the following example images, disparity won't work well on the 2D scenes but will work better on 2.5D scenes depending on the difference in extrusion and disparity between objects.

2D Scenes 

The following scenes will not produce disparity data. 

Frame_3_2x.png

2.5D Scenes

The following scenes are examples of 2.5D scenes that may or may not produce disparity. Don’t worry, the color scan image will still work in these cases.

Frame_3.1_2x.png

Frame_3.2_2x.png

3D Scenes

The following scenes will produce good disparity data.

Frame_3.3_2x.png

There are a few things that you can try to achieve better disparity results:

  • Move your projector/Lightform device closer to the scene
  • Increase the distance between your projector and your Lightform device (this may require a different mounting technique)
Was this article helpful?
18 out of 23 found this helpful

Sign in to make a suggestion to improve this article

  • Comment actions Permalink
    Avatar
    Andrew Kinsler

    The last part of the page mentions "Increase the distance between your projector and your LF1", but does not elaborate on what type of separation should happen.

    Can you give more guidance? Does this mean a vertical separation, while keeping the LF1 camera lens and the projector lens in the same plane? Same spherical plane? Should there be horizontal or Z-plane separation as well?

  • Comment actions Permalink
    Avatar
    Brittany Factura

    Hi Andrew,

    So sorry for the vagueness. We aim to update this article with more details and information.

    This means a horizontal separation. You can move the LF1 left or right in the horizontal direction from the projector, keeping the camera lens and projector lens in the same horizontal plane and height. No need for a vertical or z-plane separation.

    (For some detail, the structured light scan can be thought of in a similar way as stereo camera calibration and reconstruction (https://en.wikipedia.org/wiki/Computer_stereo_vision) and we apply similar techniques to extract disparity from the scene. The LF1 and projector can be thought of as the right eye/left eye, so increasing the distance between the two can result in more accurate disparity.)

    Here's an example of this. We used a magic arm to mount the LF1 to a separate tripod:

    Hope that helps. Please me know if you have any questions.

    Brittany

  • Comment actions Permalink
    Avatar
    John Meehan

    Hello, 

    A couple of questions -

    1. The "2D scene" above is a flat poster of the street scene, right? No disparity data because no 3D protrusions, but can this (or another flat poster) still be used as a subject? Or are the lines too thin and not enough contrast?

    2. How do you know when the disparity is insufficient, and the LF1 should be moved laterally with respect to the projector? This disparity data is used as an internal calibration for the LF1, right? How does the user know when it needs to be improved?

  • Comment actions Permalink
    Avatar
    Bruce du

    First of all, this is not a street view. This is an exhibition hall. The wall with hollow letters on the left is the actual wall of the exhibition hall, while the middle part goes all the way to the right. This is a large hand-painted painting, including street lights, people, roads, buildings, etc., which are all the contents of a plane painting. Because it's a 2D drawing, it can't produce disparity data. This is my understanding of the official meaning of lightform. If the official staff of lightform confirm that my idea is correct, I will be very happy.

  • Comment actions Permalink
    Avatar
    Ashley

    Hi Bruce, 

    Yes-- you are correct. The examples we have for 2D are flat pieces of art (a drawing on a wall, or a painting of a street) so it will not produce disparity data. I will look into using different images to convey this as I can see how those examples can seem misleading. 

    Thanks,

     

    Ashley