I'm trying to develop a tool that given a video (without a reference video), it can measure the quality. Basically given some thresholds, it assigns them a rating (I just need 4 classes 1-4). Since we don't have the original reference videos, I'm not looking for approaches using SSIM or PSNR or such. Possible options are using machine learning approaches.

Is there any available tool or package for that? If not, how can I implement one?

  • If I understand your question correctly this github.com/Netflix/vmaf could help
    – Marware
    May 24 '19 at 20:44
  • @Marware But that one is still reference-based. Isn't it?
    – Tina J
    May 24 '19 at 23:00
  • reference_path and distorted_path are the paths to the reference and distorted video files in ./run_vmaf format width height reference_path distorted_path [--out-fmt output_format]
    – Tina J
    May 24 '19 at 23:02

What you are looking for is called “no-reference quality assessment”. Google can give you specific examples.

  • Couldn't find an available tool for that, except some research papers.
    – Tina J
    May 22 '19 at 19:11
  • Which answers your question as reading those research papers will allow you to implement your own tool. @TinaJ ¯\_(ツ)_/¯ +1
    – Fabby
    May 23 '19 at 18:36
  • Haha I'm no more a phd student to spend time on that! Would prefer a ready-to-use tools.
    – Tina J
    May 23 '19 at 19:54

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.