Is the depth image returned by Microsoft Kinect SDK already undistorted?

Question

Supposedly the Microsoft SDK has access to the Kinect's intrinsic parameters but does anyone have any idea if the depth image it returns is actually undistorted? I couldn't find anything relevant.

Let me know if I'm out of topic although I consider this as an implicit programming question :)

edit: some other useful links I found that support @Coeffect's answer

• http://ros.org/wiki/kinect_calibration/technical
• Accuracy Analysis of Kinect Depth Data, by K.Khoshelham

Loyola

What do you mean by distorted? – Guidance

The IR camera of the Kinect, like the RGB camera, has its own tangential and radial distortion. I am wondering if the SDK corrects the image before it returns it. – Loyola

@downvoter: Where is the problem? What should I change? – Loyola

Hmmm.... You might want to make your question clearer, and perhaps provide some examples of distorted images – Guidance

The distortion coefficients are so small that the distortion is not visible, it's not that it's like a fish eye distortion for example. However, for my project I need as much accuracy as possible. Every camera is described by its camera matrix (focal length, principal point) and its distortion vector (tangential and radial). This is basic stuff and I feel like I would bloat up the question, especially since I tagged it camera-calibration. – Loyola

Just curious are you trying to achieve a point cloud or what? – Guidance

please meet me in the Kinect Development Chatroom – Guidance

@OutlawLemur Please stop making edits to every question you see. Those tags were not needed, and the edits did nothing productive to the question. And don't give me the "I needed to edit it to vote it up again," because you need to stop voting things down prematurely. – Ogdon

@Ogdon Those tags were not needed, but they do make the question more clear. Also the question had many needless details, and I downvoted the question initially because A) It was not clear and B)There were no examples/what there was trying to accomplish – Guidance

@OutlawLemur Tags aren't meant to help explain the content of the question, they're used to help people find questions they're interested in. No one is going to search for a "distortion" tag, or the "RGB" tag you added to a previous question. "Needless details" is just him being a human being; you don't need to minimize every post on this place. – Ogdon

@Ogdon I look at the rgb and argb tags all the time – Guidance

Answer 1

So the IR pattern that the Kinect displays isn't your normal grid. For an example, check out this blog post. The Kinect handles making a normal depth map out of this. Thinking about focal lengths and such for this system is just going to dig yourself a hole. Your thoughts about precision are probably misplaced. The Kinect isn't accurate enough to BE picky about such things. Having used the Kinect for motion detection, there's a lot of noise. If you have a certain situation in mind, you might want to post about it.

edit: Here's a post showing that the depth isn't linear, and more of the precision is focused on closer objects. So the farther away you are, the less precise the data is, and the more severe the noise will become (because having the returned depth change by 1 nearby is pretty much nothing, but farther away that accounts for a larger distance change).

Ogdon

Nice link. The ultimate goal is registration from 2 Kinects and I notice that the registration results are a little worse when the objects are further away, so I figured I should investigate distortion first. My result evaluation is empirical though, so I could be wrong here. Thanks, I'll look a little bit more to it and I'll post back. – Loyola

@Bob Agh, I know there's a post somewhere around the internet that shows that the Kinect's reported depth doesn't actually scale linearly with distance. It becomes less and less accurate as you move farther away. If I can find the post, I'll update the answer. – Ogdon

Thanks, I'll go with that. If anybody else is interested, I also found this: ros.org/wiki/kinect_calibration/technical – Loyola