simple image clustering
in [Matlab]
|
From: uny gg on 29 Mar 2010 16:22 I would like to apply simple image clustering methods such as K-means or hierarchical clustering. However, the problem is my image size is very large. For example, most of my image size is 1.5 M and the dimensions are around 5000 by 4000 pixels. So, if I handle in pixel levels, it is almost impossible to handle it. I am not sure what kind of technique I can use for this. Could somebody please give any suggestions?
From: ImageAnalyst on 29 Mar 2010 19:53 1.5 MB must be the compressed size of your image because an uncompressed 4000 by 5000 color image would be 20 MPixels or 60 MB. Still, even 60 MB is not large in general. And 20 MB is only somewhat larger than a garden variety digital camera these days. Maybe it's too big to do kmeans on - I don't know - but if that's the case why don't you just block average your image down with imresize()? Chances are you're just looking for a few clusters anyway (say less than a dozen) so you don't need that extreme resolution. You should be able to find your major clusters (the biggest 20 or so) just fine with the resolution cut down by 1/8 or 1/16 or whatever. If you don't believe me, then post your image to prove it to us.
From: uny gg on 30 Mar 2010 12:37 Thank you ImageAnalyst!! Original color images size were 2M to 6M.. they were various.. However, when I transform into the B/w images, the size was decreased.. ;) So, now most of images are 1.5MB but dimensions are still around 4000 by 5000 pixels. As you suggested, I have applied imresize() function to reduce the dimension so, I can handle in a pixel level. thank you for your suggestions. ^^
|
Pages: 1 Prev: parfor with sedumi or yalmip Next: Using a for loop with ode45, 2nd order ODE |