I am working to identify and isolate each human limb of a known person, in pose, in an image which is relatively noisy. It should also be noted that the image is a snapshot extracted from a video (which can be utilized in this task).
The end result ideally would be a bunch of polygon estimates of each body limb segment (i.e. torso, head, neck, upper arm, lower arm, hand, upper leg (thigh), lower leg, and feet/shoe). I currently have results that allow me to find the contours of the image, though these include materials that are prevalent in the background (i.e. its not a clean image of only a human).


1). I am currently using opencv to implement the task, so any answers that can be directed towards that API are greatly appreciated, though I will happily accept any general solutions that answer the question. 2). If you feel the question is lacking in clarity, please comment as to why. I truly feel this question would add value to the community if answered completely.


How can I create polygon estimates of each body part/limb that is present in an image which always contains one person?

  • $\begingroup$ The current state of the art in finding body parts is research.microsoft.com/apps/pubs/default.aspx?id=145347, but starts with depth images. The paper has a good related work section though, which will give you a good reading list to start. $\endgroup$ – Wandering Logic Jun 2 '13 at 12:57
  • $\begingroup$ Are the parts labelled ? If so this becomes a polygon fitting problem on binary shapes. $\endgroup$ – beedot Jan 6 '14 at 2:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.