I want to probabilistically infer the binary state of nearby pixels from some samples. Example:
? ? ? ? ?
? ? 1 ? ?
0 ? ? 1 ?
? ? ? 1 ?
? ? ? ? ?
I have a few examples to measure the spatial autocorrelation in this type of data; the spatial structure of new images should be similar (it represents lakes and rivers). example map How can I use e.g. Moran's I to infer the probability of the "?" pixels in the example? Does Moran's I even provide useful information about pixels more than 1 unit apart from the sample? Can I realistically get good probabilistic expectations for points a few units away from the nearest sample based on auto-correlation or is it the wrong tool for this job?

  • 1
    $\begingroup$ Possible direction: Image restoration technique using Kriging with combination with indicator kriging $\endgroup$ – Dahn Mar 20 '18 at 10:19
  • $\begingroup$ I have some doubts about the viability of this algorithm with few known pixels that are spread out relatively far, but I will look into it. Thanks for the idea! $\endgroup$ – sh4dow Mar 21 '18 at 16:30
  • $\begingroup$ I concur with Dahn Jahn (above). This is an interpolation problem and it sounds like indicator kriging would be your best bet. $\endgroup$ – coreydevinanderson Apr 6 '18 at 3:18

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.