3
$\begingroup$

The activity going on in this thread, I do not think it can be constituted as mathematics. It think it is image processing, but still not the mathematical aspect of the image processing but purely on technology needed to do something on medical images.

IMHO this would best be discussed on signal processing stack exchange. Please share your opinion.

$\endgroup$
  • 5
    $\begingroup$ The question you refer to is not formulated as a mathematical question, and it is not clear to me what the precise question actually is -- "similarity / dissimilarity" of images is neither a standard mathematical term, nor is it defined in the question. To me the question looks like a kind of "fishing expedition". $\endgroup$ – Stefan Kohl Jun 14 '15 at 16:42
  • $\begingroup$ @stefan yes including the question. $\endgroup$ – Rajesh Dachiraju Jun 15 '15 at 2:17
  • 5
    $\begingroup$ if the question is focused onto the most common medical application, image alignment, then there is actually a quite sizable mathematics literature; here is an overview --- siam.org/meetings/op08/Modersitzki.pdf $\endgroup$ – Carlo Beenakker Jun 15 '15 at 6:34
  • 3
    $\begingroup$ In addition, computer vision and pattern theory do have a lot of underlying mathematical theory and have attracted some number of mathematicians. There are especially a lot of recent effort in applying differential/Riemannian/Finlerian geometrical techniques to the classification and similarity and dissimilarity between curves (e.g. this recent review paper of Michor and Mumford) and extension to higher dimensional objects (e.g. this paper).... $\endgroup$ – Willie Wong Jun 15 '15 at 10:27
  • 8
    $\begingroup$ ... In short, while there may be things wrong with the question (mostly having to do with imprecision of the phrasing), a generous interpretation of the question certainly can be about (not necessarily applied) mathematics. I should think the question maybe "too broad" or "unclear", but until it narrows down I cannot justify it being "off-topic". $\endgroup$ – Willie Wong Jun 15 '15 at 10:30
  • $\begingroup$ @Willie : There is this thing especially about applied signal processing and computer vision problems is that there is a considerable and non-trivial gap between the problem statement/application name (usually this is physical description of problem) and then the later one being a mathematical formulation of sub problems which could be probably solved by using mathematics. I mean to say, just modeling being allowed doesn't mean that one can ask "I want to do alignment of medical images", as it is just a crude problem statement without any mathematical formulation. $\endgroup$ – Rajesh Dachiraju Jun 15 '15 at 10:36
4
$\begingroup$

From a quick look, it seems a modelling problem. It's basically asking "what is the best function with this properties". Where the meaning of 'best' is far from clear. Essentially, a function that does what I want with the data I have. It's not a technology problem.

It could be argued that this specific question needs to be closed with both the too broad and unclear what you are asking reasons, but a better point to discuss here is whether modelling questions are ok on MO.

People have discussed this topic at math.stackexchange, and proposed to open a new SE site (the proposal wasn't successful and has now been deleted).

I see three possible answers:

  1. This kind of questions is off-topic here. Until there is no mathmodelling.stackexchange.com, they live in the "stack exchange limbo" where there is no SE site available for them.

  2. This kind of question is officially on topic here. We don't want yet another math-related SE site, and the borders between the two disciplines are blurry anyway. Let's modify the site policies.

  3. This kind of questions is a better fit for http://scicomp.stackexchange.com, because they are more linked to computational/applied side of mathematics than to the pure discipline. We can migrate them manually. If compsci.SE (which is still in beta) closes, goto 1.

It is ultimately a political decision, "what do we want this site to be". All choices have their pros and cons. My understanding is that the common consensus among the Mathoverflow moderators and high-rep users is more towards 3.

$\endgroup$
  • 3
    $\begingroup$ In this connection it might be instructive to consider the development of Compressed Sensing, and the role that Terence Tao played in this application of pure math to data analysis. Can you imagine that we would have closed this as "off-topic" for MO? --- terrytao.wordpress.com/tag/compressed-sensing $\endgroup$ – Carlo Beenakker Jun 15 '15 at 8:18
  • 3
    $\begingroup$ Why this strict dichotomy between "modelling questions are on-topic on MO" and "modelling questions are off-topic on MO"? -- Can't we say that particularly good such questions are welcome here, but we don't want a huge bulk of such questions -- in a similar way as with soft questions? $\endgroup$ – Stefan Kohl Jun 15 '15 at 8:42
  • 1
    $\begingroup$ @StefanKohl I see the dichotomy more as "modelling questions are off-topic per se, in all cases, just because of what they are" versus "modelling questions are not forbidden, but of course they still have to satisfy the same criteria as the other questions to be acceptable, especially if they are 'soft questions' without a well-defined answer". $\endgroup$ – Federico Poloni Jun 15 '15 at 9:01

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .