Yes you can train a classifier to detect and locate object(s) in an image. Object category detection and classification is an exciting area.                                                   

Take a look at the following link: http://www.image-net.org/challenges/LSVRC/

List of good papers:    http://image-net.org/about-publication

You need to define better the scope of your application. First, what kind of objects are you aiming to detect. From the answer to this question, you need to define the image decriptors to use (edges, colors, textures, etc). Then you can use such features as an input to a SVM or any other trained classifier. 

For instance you can look for papers related to pedestrian detection, where a SVM is used, with Histogram of Gradients (HOGs) for human shape detection.

This page is an excellent resource:

http://www.cs.berkeley.edu/~rbg/latent/

Thanks M Raza Ali, Resources are really useful... if more links you know..can you please share?

Saul Calderón Ramírez, My images are medical images...

I think the papers will be a good starting point to give you a comprehensive background.

This website will give you a good overview and provides download links for extracted features:     http://www.cloudcv.org/

also checkout this page http://objdetect.cloudcv.org/trainaclass/

Thanks Kouser and M Raza Ali..