Is it easier to stitch 1080@30fps in h.264 or MPEG?

asked 2016-09-25 19:13:52 -0600

olibd gravatar image

Hello, I am a software engineering student and for our final project we need to do live video stiching using OpenCV, we intend to run the software on a laptop with a graphic card that supports CUDA.

From a processing and technical standpoint, would it be easier for to stich images coming from 3 cameras that output 1080x1920 images at 30fps in MPEG for a total bitrate around 180Mbps or would it be easier to stich the images coming from 5 cameras that output 1080x1920 images at 30fps in H.264 for a total bitrate around 60Mbps?

Thank you!

edit retag flag offensive close merge delete

Comments

2

IMHO when you are doing OpenCV operations, you work with uncompressed data stored in Mat structure. So compression won't have any impact on processing time.

I imagine you'll be using fixed cameras, so you can do the stitching with precomputed transformation matrices (otherwise it will be difficult to process in real-time 5x2Mpixel images).

The stream decoding is done by hardware, the question is if your laptop can decode 5 full-hd H.264 streams. I never tried, so I have no idea.

kbarni gravatar imagekbarni ( 2016-09-26 02:25:38 -0600 )edit
1

I will note that 3x1080 is almost certainly easier than 5x1080 as far as stitching goes.

Tetragramm gravatar imageTetragramm ( 2016-09-26 20:31:19 -0600 )edit