php - Find most favored area on an image -


what mean-stat-equation should use when have image n-number sample-size of selections?

i have unique problem hoping advice, don't miss out on anything. problem: find favored/liked/important area on image based on user selection of areas in different selection ratios.

scenario: consider image of dog, , hundreds of users selecting area on image in various resolutions, obvious area of focus in selections area containing dog. can record x1,x2,y1,y2 co-ordinates , put them db, if want automatically generate versions of image in set of resolutions should able recognize area max attraction of users. methods think work are:

  1. find average center point of selections , base selection in that. - simple not accurate.
  2. use algorithm k means or em clustering don't know 1 best suited.

looking forward brilliant solution problem

more info on problem: actual image 1024x768 image, , selections made on of common mobile phone resolutions. objective automatically generate mobile phone wallpapers intelligent learning based on user selections.

i believe have two distinct problems identified above:

one: identification of points

for this, need develop sort of heuristic identifying whether point should considered or not.

i believe mentioned hundreds of users selection locations on image? hundreds may lot of points cluster. consider excluding outliers (by removing points not have number of neighbors within particular distance)

anything can reduce dataset helpful.


two: clustering of points

i believe k means clustering best suited particular problem. link

your particular problem seems closely mirror standard cartesian coordinate clustering examples used in explaining algorithm.

what you're trying appears np-hard, should satisfied classical approximations.

once clustered, can take average of points within cluster rather accurate approximation.

in addition:

you dataset sounds tightly clustered. (i.e. people pick dog's face, not side of it's torso.) need aware of local minima. link these can throw wrench algorithm. small number of clusters. aware may need bit of dynamic programming combat this. can introduce variance algorithm, allowing average points "pop out" of these local minima. local minima/maxima

hope helps!


Comments

Popular posts from this blog

java - Play! framework 2.0: How to display multiple image? -

gmail - Is there any documentation for read-only access to the Google Contacts API? -

php - Controller/JToolBar not working in Joomla 2.5 -