I noticed that the IR camera, and therefore the depth stream, has a slightly barrel distortion. This is of course more noticeable on the edges of the image.
Upon further investigation, I noticed that the reported internal camera parameters of the IR camera have less precision than the camera parameters of the color camera. For example, the principal point of the depth camera is always exactly w/2 and h/2, whereas the principal point in the color image is an actual float value with more precision. The same holds true for horizontal and vertical fields of view.
After I obtain the IR image from the camera, I can correct the barrel distortion, and calculate better internal camera parameters, but the depth stream of course continues using the initial IR data as the source.
So, how can I get corrected depth images?
I think there are two ways of proceeding here:
1) is there any way to generate a depth image from my own modified IR image?.
2) Is there any way to modify the internal camera parameters of the IR camera in the SDK?
