how to use feature vectorization and pca for feature reduction?

I have already extracted multiple features like color, texture, shape extracted using methods and algorithms like regionprops, color moments, glcm, vein feature and so on. I have 110 features per image. kindly guide me how to form feature vectorization and use pca for feature reduction.

thnx

Answer 1

[Image Analyst]

Basically you've asked us to give you a course in Image Processing in an Answers forum posting. So here it is : http://szeliski.org/Book/

I'm also attaching my pca demo, actually given to me by the Mathworks.

[parul mittal]

thanx sr for reply. sr I have already extracted features. now I have formed feature vector of size 16*110 also where 16 are images and 110 are features/per image. I just want to know about pca. kindly guide me.

thanx

[Image Analyst]

Well I gave you a demo. And the help has more. Why can't you just pass your data in?

coeff = pca(X)

If you still don't know what to do then attach your feature matrix in a .mat file and I'll pass it in to pca() for you and give you coeffs, though you could of course do that yourself.

[parul mittal]

thanx sr.. yes sr I can compute. sr can u guide me can I directly use this output as an input to classifier or i need to normalize it? and sr I m bit confused whether coefficients means matrix of elements of eigenvectors and scores means eigenvector or eigenvector*data? kindly guide me.

thanx

[Walter Roberson]

It is not required that you normalize features. However, normalizing features might give you much better results. Why not try it both ways?

[parul mittal]

thnx sr il try in both way