Monochrome Capture
Introduction
The Zivid 3D cameras allow subsampling on the camera side. We call this monochrome capture and it provides the following benefits:
- Speed
Less data to transfer and process leads to faster captures.
- Narrow band
Improved ambient light resiliency.
- Increased distance between pixels
Reduced Contrast Distortion and highlight effects.
Resolution
The following illustration shows the bayer filter grid and associated indices.
When we only use e.g. the ‘Blue’ pixels we get 1/4 of the original number of points.
This means that, after subsampling 2x2, the number of points along x and y axis is divided by 2. For subsample 4x4 the number of points along x and y axis are is divided by 4.
For example, for Zivid 2+ M130 the spatial distance between two pixels is 0.32 mm @ 1300 mm working distance. With 2x2 subsampling, this distance is doubled to 0.64 mm at the same working distance.
This does not affect the resolution in depth; the point precision determines this.
2D resolution
Full Resolution - 2D
Quarter Resolution - 2D
3D resolution
Full Resolution - 3D
Quarter Resolution - 3D
Speed
TL;DR: Subsampling with monochrome capture provides significant speed improvements.
By reading out only a fraction of the pixels there is less data to transfer and less data to process to generate a point cloud. Additionally, hardware subsampling eliminates the need to downsample the data to transform it to a more manageable size and thus reduces the storage and post-processing requirements. With a quarter of the data, Zivid post-processing (such as Normals and Transform) and user post-processing (e.g., CAD matching) become faster.
Quality
Regarding 3D quality there are in particular two improvements with monochrome capture.
- Narrow band
Improved ambient light resiliency.
- Increased distance between pixels
Reduced Contrast Distortion and highlight effects.
The increased distance between pixels reduces the amount of signal leakage from neighboring pixels. This is the primary source of the two negative effects:
Mapping between a full-resolution 2D image and a subsampled point cloud
In many applications, it is common to capture a 3D point cloud separately using Monochrome capture, along with a 2D image using a dedicated 2D capture.
This approach offers several benefits.
It allows for independent optimization of the 2D settings for the image and the 3D settings for the point cloud.
By disabling color data processing with
Sampling::Color::disabled, the processing to get the point cloud becomes faster.It is possible to capture 3D while processing 2D, and vice versa.
It is possible to maintain a 1-to-1 correspondence between the 2D image and the 3D point cloud. This can be achieved either through resampling or by mapping between the pixel indices of the 2D image and the 3D point cloud.
Version History
SDK |
Changes |
|---|---|
2.12.0 |
Full resolution captures are up to 10% faster and subsampled captures are up to 20% faster with colors enabled on low-end GPUs. |
2.11.0 |
|