Quantcast
Channel: Raspberry Pi Forums
Viewing all articles
Browse latest Browse all 7995

Camera board • Re: What is faster taking a small picture quickly or video stream capture_array?

$
0
0
Okay so I made some changes/tried things out.

What I wanted initially is as shown in my attachment, I wanted 3 zooms: 1x, 3x, 5x with the resolution multiples being based on the display eg. 640x480.

When I tried to make the video configuration for the 5x zoom eg.

Code:

self.video_config_1x...self.video_config_3x...self.video_config_5x = self.picam2.create_video_configuration(    main={"size": (3200, 2400), "format":"RGB888"})
It would not work, initially I thought it was a limitation on the number of times you can call

Code:

create_video_configuration
. I would see this message "no such process" when picamera2 would try to start with that config/crash.

I did look at that "camera zoomed in" thread where it shows the resolutions eg.

Code:

2028x1520 [40.01 fps - (0, 0)/4056x3040 crop]
If going higher than that results in 10fps then I'll just stick with that.

Although in order to achieve my 5x zoom, I just upscaled the

Code:

2028, 1080
resolution. Which I initially had 1920x1440, I'm not sure if it actually was using that 2028 combo.

Anyway I'm using cv2.resize eg. width * 1.58, height * 2.22 to get the 5x. I have to rework the math on the resolution/panning since I initially did the multiples of 640x480 but it seems I need to use that 2028x1080

Using MJPEGEncoder now, using cv2.imdecode

I looked into ScalerCrop where it says:
A libcamera.Rectangle consisting of:
x_offset
y_offset
width
height
I don't get how that's different than calling cv2 crop, maybe performance is better. Or is it taking a high res video and then outputting that bit only so save framerate by not having to use cv2. Alright I'll try that again. The math will change... panning then upscaling...
zoom-plan.png

Statistics: Posted by jcun4128 — Thu Feb 19, 2026 3:15 am



Viewing all articles
Browse latest Browse all 7995

Trending Articles