1

I work on image and video quality assessment and I am attempting to build a very large image/video database.

Almost all of the images that are uploaded in social media websites are, if not post processed, atleast well captured. Most of us take 100s of pictures and end up uploading only a few good ones. Is there an easy way to gather the ones that are not of high quality (eg: blurry, under/overexposed)?

deeptigp
  • 373
  • 1
  • 3
  • 9
  • Are you trying to snoop on peoples photos that they choose not to share? Or are you trying to identify ones that are not of high quality (the reason why they might not be shared)? Either way you are going to need people who acknowledge what you are doing and choose to install your software on their phone. – dhj Jun 27 '14 at 02:55
  • I definitely want people to approve that their pictures will be used for academic research. All that I am looking for is a source to gather more pictures, which have one or more distortions ( noise, compression, blur etc.) – deeptigp Jun 27 '14 at 06:52
  • Hmmm... As you probably know, there are widely available photo sharing sites -- but like you said, those are going to be self selecting less noise / compression / blur / etc. If you're trying to have an algorithm identify those components the you could always add them algorithmically (gaussian blur filter, random pixel distortions, compression, etc) and see how your algorithm performs. If you want a person to identify those and still share them then you probably need a custom app which would mean development and adoption barriers. Depends on algorithmic interest or human decision interest. – dhj Jun 27 '14 at 18:20
  • There are datasets available which do just that - add a known amount of distortion to a fixed set of images. I however am shooting at gathering naturally distorted images. And ya, you are right. This might mean we would have to build a specific mobile app to cater to our needs. – deeptigp Jul 11 '14 at 15:09

0 Answers0