How do I calculate the points in a video sequence where the images are steady enough for a photo. For example, I want to take maybe 20 photos for a facial recognition dataset. Instead of asking the subject to hold still 20 times, I take a moving video of him and filter the images for 20 good photos. The problem is that the subject face has to move into different poses and facial expressions and many times the movements blur the quality of the image.

Thanks in advance.

  • $\begingroup$ If you have a video you could filter based on the optical flow between images. There are also "blurness" metrics, such as the variation of the Laplacian, that you can use to determine the "amount of blur" an image has. $\endgroup$
    – JVGD
    Jun 19 '20 at 8:43
  • $\begingroup$ Thanks. What I ended up doing was to go in straight with the face detector and filtering it for highest confidence levels. $\endgroup$ Jun 22 '20 at 1:54

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.