Danella's profile - activity

Thank you so much Eduardo! I searched a lot and couldn't find this post. Now I know what is happening.

Hi everyone,

I am working with the calibration of four cameras of a quadro camera 360 degrees setup, and I am getting some really strange results. I tried searching online about it, but it did not return me not any clear explanations of possible causes of this problem.

I am attaching some images for you so you can see it, but some of them differ a lot.

What I realized after doing some tests is that if use images which my chessboard placed near the limits of the view, the expected result (nice calibration) will not be achieved. Although when I use a few images and with the board near the center of the view, the calibration algorithm returns to me a better result. I got some nice calibrations using just 5 or 6 images, but since these results are not general (chessboard almost in the center always), it does not undistort all kinds of images that I can take (I can still see the corners a bit circular in some images that I applied the results from the calibration, instead of straight patterns).

Is there anyone who can give me a nice explanation of why this is happening? And more than that, maybe clarify what I should do in order to get a good calibration and not what I got?

Thank you!!

Image1 Image2 Image3 Image4 Image5

Okay, the next step is to unwarp the resultant images produced, and then try to stitch each one of the four images in order to create panoramas using these four cameras. Considering that they are always in the same position in relation to each other (they're mounted together) the parameters that I have gotten using the calibration would not change.

Just to be more clear: 1) Calibration (done) 2) Unwarp 3) Stitch 4) Enjoy the panoramas

Good morning everyone,

I've been trying to calibrate a 360 quadro camera using opencv's algorithm for calibration. After a few days I got nice results for each one of the four cameras, but now I am a kinda lost. I do not know exactly what should be the next step of the process of getting undistorted images for every taken image. Given that I already have the intrinsic matrices and all the parameters necessary, what would be my next task to implement these solutions to all the images that I'll be producing with the camera?

I have been learning a lot by studying OpenCV. Then, I am sorry if I am asking an obvious question or something silly.

Thank you.