Using Principle Component Analysis (PCA) in classification

Hi All, I am working in a project that classify certain texture images. I will be using Gaussian Mixture model to classify all the database into textured and non-textured images.

Now, I am using PCA to reduce the dimension of my data that is 512 dimensions, so I can train the GMM model. The results from PCA are new variables and those variables will be used in the training process:

[wcoeff,score,latent,~,explained] = pca(AllData);

The question is: in the testing process how can I use the wcoeff to get the same variables? Do I just multiply the wcoeff with the new image?

[Delsavonita Delsavonita]

i have the same problem too, since you post the question on 2014, you must be done doing your project, so can you kindly send me the solution for this problem ? i really need this...

[Adam]

Don't post your e-mail address in a public forum.

Answer 1

[KaMu]

I keep received emails that some one answer my question but I can't see any answers!

[Image Analyst]

Because we don't understand your question. See my attached PCA demo. It will show you how to get the PC components.

[jin li]

It is right. He finally display each component. first calculate coeff then component=image matrix * coeff so this will be eigenimage