0

I extracted features from video samples and then each feature is normalized (grouped/quantized/clustered) to values from 1-4 or from 1-3. I need to do my experiment in few samples that cover all features quantized values.

Example:

sample f1 f2 f3,

s1 1 2 1,

s2 2 2 1,

s3 2 1 2,

s4 1 1 1,

s5 3 1 1,

here I can choose s1, s3, and s5 to represent my samples. I can skip s2 because I have f1=2 in s3 and f2=2 in s1 and f3=1 in s1 and s5.

thanks

user3569998
  • 15
  • 5
  • Unclear. Could you edit question and elaborate a bit ? – kebs Nov 16 '15 at 22:05
  • I edited it. tell me if it is not clear. – user3569998 Nov 16 '15 at 22:21
  • Sorry, I wasn't clear enough. What tool/programming language do you use ? What do you call a "feature" in a video sample ? As such, I'd be surprised if anybody can answer. Is this more an "algorithmic" question ? – kebs Nov 17 '15 at 13:44
  • I am using matlab. The feature might be the std of the frames' Intensity. – user3569998 Nov 17 '15 at 13:49
  • Then add the `matlab` tag to attract matlab guys over here. And add your comment to the question. – kebs Nov 17 '15 at 14:06

0 Answers0