If you are following us from previous two posts (StratusNet and Video Streaming & Compression), you are now ready to understand better how our depth compression works.
The Stratus Depth Compression uses the sensor accuracy, a logarithmic transformation and the run-length encoding.
Looking into the linear mapping of the sensor shown in the following figure,
It shows that the sensor has not enough precision for close objects and has too much precision for far objects. Because of it the linear mapping requires a large numerical range (0-50000).
Performing the logarithmic transformation makes mapped data to use the sensor precision accurately both for close and far objects and it reduces the numerical range considerably, as shown below:
The Stratus Depth Compression also makes use of the run-length encoding (RLE). The Run-length Encoding is simple and lossless compression. The depth image characteristics makes the RLE a good approach, since the image has many values that are repeated.
The logarithmic transformation and RLE together:
- The log transformation makes RLE more efficient
- The numerical range reduction provides extra 5 bits to store run-length encoding counts
Thank you for your interest,
Keep following us.
