SF3325-100Camera fisheye correction

Hello,

I am using the SF3325 camera to record a screen on which is running a video, which i later have to compare with the video that ran on the screen.

The image from the camera has fisheye distortions. I have tried to flatten the image by using the provided sample application “Video Rectifier” but the results are as follow, see attached(VideoRectifier_results.jpg).

I will also attach an image from the original video and an image from the video recoded by the camera.

Could you advise what is the problem?

Many thanks.!


VideoRectifier_result|690x273

Hello @liviu.poenaru,
this is a very special case, will check try to reproduce it and check it internally.
meantime, can you please share more information:
are you using Drive Software 10 / 9?
is you screen flat? (can specify the type of screen brand?)
what distance are you putting the camera?
and what are you trying to achieve with the comparison?

thanks

Hello,

Thank you for the quick answer.

I am using Drive Software 10

The screen that was recorded is a TV model:
Sony BRAVIA KD-55XF8505B
Screen diagonal: 139cm
Screen size: 55 inch
Aspect ratio: 16:9
Resolution: 3840x2160 (UHD)
Backlight: LCD

The approximate distance is 1.1m, which has been chosen so that, with the default focus of the camera, the image would look sharp.

By comparing the 2 videos, i am trying to:

  1. Analyze if the object tracking (cars) in the 2 videos is similar => the same objects are tracked and for the same duration
  2. Assess the delay introduced by the camera

Hi liviu.poenaru,

Can you share how you are capturing/recording the video from the SF3325 camera pointed at the screen?

Also wondering if there is a particular why you are tracking the two HW flows? i.e. (video from file -> object tracking vs video from file -> Screen -> Camera -> object tracking)? Typically it’s much more straight forward/accurate/useful to just do object detection directly from camera or from video from file, so want to make sure you end up with results you are after. Are you expecting the two images to be near identical with this flow?

Hi @liviu.poenaru,
please update if this topic is still relevant or can be closed.

thanks