Page 1 of 1

Vertical decimation

Posted: Sun Feb 16, 2020 4:05 pm
by rahveegmail
When doing robotics and depth map, decimation is useful to preserve the full FOV, but I'd much prefer vertical decimation, to preserve full horizontal resolution (preserving depth precision). I prefer to sacrifice vertical resolution instead of horizontal. I can't find any option to do this. Am I missing it? Or does it not exist? Would it be easy to add in software?

Re: Vertical decimation

Posted: Mon Feb 17, 2020 9:55 am
by Realizator
Not all stereoscopic features are implemented in PiCamera correctly (but most of them works fine with raspistill/raspivid), and top-bottom mode is unstable.
As a solution, you might exclude decimation at all.
For example, you need to get two images with 320x240 from each camera.
If you use decimation, you get one stereopair with the 320x240 resolution, where each camera gives you 160x240 image. So each image is scaled by 0.5 horizontally.
But you can request 640x240 with no decimation. So you have 320x240 for each camera, with no decimation. We use this approach in all our latest code (here for example: ... /