I am trying to reduce the dimensions on an a set of images using Matlab Toolbox for Dimensionality Reduction. Problem is: I know very little about dimension reduction. So am trying each one by trial and error, passing the data set to the function. I have tried 6 so far, PCA was returning a matrix with a complex number. And the others was frozen matlab. What Image reduction methods is suitable for images?
Asked
Active
Viewed 3,206 times
4
-
Off-topic: Would be better-suited at http://dsp.stackexchange.com. – Oliver Charlesworth May 07 '12 at 16:41
-
1Why do say that? Cause I was wondering if it better suited for Cross Vaildated – cubearth May 07 '12 at 16:50
-
Because your question seems to be closer to "how does PCA work?" rather than a programming question. – Oliver Charlesworth May 07 '12 at 17:29
-
I really meant to ask. Which dimension reduction method should I use on images... Will edit to be more actually. – cubearth May 07 '12 at 18:01
-
What do you aim via dimensionality reduction? PCA is good for maintaining reconstruction error (that is the best low rank approximaton in terms of l2 reconstruction error), but is that really what you want? – petrichor May 07 '12 at 18:08
-
Trying to recognize patterns on clothes – cubearth May 07 '12 at 19:12
1 Answers
3
It looks like you are trying to implement something like eigenfaces, so first of all take a look at the article. Normally you need to get eigenvectors either with PCA or with SVD. However, if you need really low resources use, take a look at Random Projection method.
ffriend
- 27,562
- 13
- 91
- 132
-
1Here note, that SVD may be used as one possible implementation of PCA, but PCA has other implementations too and I don't know which on is used in Matlab toolbox. – ffriend May 07 '12 at 21:31