Do the image pairs satisfy the standard form?

Software
1

Hi! I’ m using a ZED 2i and ZED SDK v4.0.8. I use the following code to get image pairs from a pre-recorded .svo file:

image = sl.Mat()
while True:
    if cam.grab(runtime_parameters) == sl.ERROR_CODE.SUCCESS:
        cam.retrieve_image(image, sl.VIEW.SIDE_BY_SIDE)

Then I want to try new stereo correspondence algorithm to get disparity map instead of DEPTH_MODE in SDK. We know the image pair need to satisfy standard form (corresponding points are constrained on the same image scanline) before stereo matching. It can be described with the following picture

image
image825×605 40.7 KB

I want to know if the image pairs obtained using the SDK satisfy this condition? Do I need to do additional correction of the image pairs?

2

I tested it with a checkerboard. It seems that not all pixel points satisfy this condition.

video_0000_0330
video_0000_03301920×540 58.1 KB

We can see from the first and second red lines in the image that the vertices of the black grid are not aligned well. But the vertices of the black grid on the third red line are aligned perfect. It seems that the pixels near the left side of the image are not aligned. Is that reasonable?

3

Hi @niuge12138,

The SIDE_BY_SIDE parameter does give out rectified images, further correction should not be needed on your part.
Slight misalignments may appear on the sides of the pictures. You can try setting the camera_disable_self_calib InitParameter to true, it might improve the results.