1. Home
  2. EF EVE™ VolCapp
  3. Tutorials
  4. Step 5: Volumetric Capture Pointcloud Filters

Step 5: Volumetric Capture Pointcloud Filters

For volumetric capture you need to apply the pointcloud filters. Depth sensors are not very precise, therefore they produce a lot of inaccurate depth data. These inaccuracies need to be completely cleaned in order to have a correct volumetric video.

In simple terms, your goal is to record only the most accurate depth data. Don’t be afraid of the changes the filters do – it will look like you’re deleting a lot of important data, but this is the only way to capture a good volumetric video.

Pointcloud filters are applied to clean it and get ready for volumetric capture
Raw pointcloud without any filters and filtered pointcloud

Depth cameras produce depth data which decreases quality regarding the distance

Pointcloud quality differences regarding distance from depth camera
Distance from depth camera influences pointcloud quality

Applying the pointcloud filters

To get the best view possible, you also need to adjust the pointcloud filters. These filters can sharpen your capture, create smoother edges or surface. 

Examples are done using Kinect Azure depth camera. With Kinect 2.0 or Intel Realsense the results may vary.

  • To make these changes, scroll down to the Pointcloud Filters section and click the checkbox enable filters
  • To adjust the intensity of filters move the slide bar to the left or right:
EF EVE VolCapp pointcloud filters enabled for volumetric capture
Pointcloud filters enabled for volumetric capture

The most important filters are Noise Reduction and Steep Depth Removal. We will now lead you through the steps to apply these filters and later on, we will show you the rest.

  • Firstly, in Calibration tab, you will need to scroll down to the Pointcloud Filters section and click Enable Filters:
Pointcloud filters for volumetric capture at EF EVE VolCapp
Enabling the pointcloud filters on EF EVE™ VolCapp

If you turn on the 2D view while the pointcloud filters are applied, you will also see less data there, because the filters clear a lot of data. The applied filters will stay next time you open EF EVE VolCapp, so if your 2D view is not full, check if the filters are turned off.

Noise reduction filter

Cleans randomly positioned points in small groups. You can adjust the density of the group of these points, distance, and how many iterations (times) noise reduction filter will be applied. 

  • Zoom in your 3D view.
  • Enable the Noise Reduction filter
Volumetric capture, noise reduction filter
Noise reduction pointcloud filter for volumetric capture not applied yet
  • Slide the Density, Distance and Iterations bars to the biggest value at the right:
Noise reduction filter applied for volumetric capture
Noise reduction pointcloud filter applied for volumetric capture

See the Noise reduction filter in action:

Volumetric capture pointcloud filter noise reduction EF EVE VolCapp
Noise reduction pointcloud filter cleans the unnecessary data from your volumetric capture

Steep Depth Pointcloud Removal

Surfaces facing depth camera are captured very well. However, surfaces that go from the edges forward is very inaccurate. That’s why Steep Depth Removal is the most important filter of them all.

Kinect Azure depth accuracy scheme
Kinect Azure depth accuracy – red surfaces depth is not very accurate

When you apply the Steep Depth Removal filter to your volumetric capture, it may look like it cleans a dramatic amount of data, but actually the removed data is too inaccurate to create a good volumetric video. We recommend to use this filter at 100% in most of the cases.

  • Scroll down in the Pointcloud Filters section and turn on the Steep Depth Pointcloud Removal filter:
Steep depth removal filter applying for volumetric caprure
Depth camera pointcloud filter steep depth removal not yet applied
  • Move the Intensity slide bar to 100%
pointcloud filtering steep depth removal EF EVE VolCapp
Steep depth removal pointcloud filter applied for volumetric capture

This is the Steep Depth Pointcloud removal filter in action:

Volumetric capture differences with and without steep depth removal pointcloud filter
Steep depth removal pointcloud filter aplication for volumetric capture

Both of these filters need to be used to get the best results for your volumetric capture. Noise Reduction and Steep Depth Removal filters are the most important ones.

Multiple Filters Applied

You can apply more than one filter to get the most realistic volumetric capture. There are various possible filter combinations that you can apply. You need to decide which filters you will use by trying each one of them.

Pointcloud filtering applied cleaning volumetric capture
Pointcloud filters cleaning the volumetric capture from incorrect data

All filters use the PC’s CPU – the more filters you apply, the more power it takes. You can check the CPU in Windows Task manager

We have selected three sets of filters that we suggest to use:

  • Noise Reduction + Edge Smoothing + Steep Depth Removal
Volumetric capture quality with pointcloud filters applied
First pointcloud filters set applied on EF EVE™ VolCapp for a better volumetric capture quality
  • Edge Smoothing + Surface Smoothing + Steep Depth Removal
Better pointcloud quality while using pointcloud filters for volumetric capture
Second pointcloud filters set on EF EVE™ VolCapp for better volumetric capture
  • Outline Removal + Steep Depth Removal
Pointcloud filter steep depth removal used alone
Steep Depth Removal pointcloud filter only
Pointcloud filter set makes a better volumetric capture than using only one filter
Third pointcloud filter set – Outline Removal and Steep Depth Removal used together

Removing the floor

If you want to upload your volumetric capture into a volumetric environment, you may want to remove the floor from your capture. This is done in the Bounding box section.

  • Scroll to the Bounding box section and click the + sign. It will open your options to adjust the bounding box:
EF EVE VolCapp bounding box enabled for volumetric capture
Bounding box enabled for volumetric capture
  • You can adjust the X, Y, Z axes, and in the small squares next to them, the sides of the bounding box are marked.
EF EVE VolCapp bounding box calibration
Bounding box calibration sliders for volumetric capture
  • Select the Y axe, with a marked bottom of the square and raise the number up until the floor disappears:
Removing floor from your volumetric capture
Raising the y axe to remove floor from your volumetric capture

It’s best to have a person or an item in the capture stage, then you won’t remove too much floor. Otherwise it might cut the feet from the view.

Read next how adjust the depth camera colors

Updated on January 31, 2020

Was this article helpful?

Related Articles

Leave a Comment