I have been trying to implement image classification by extracting facial features. I tried starting with some already established methods like Multi Kernel Approach, using Independent Component Analysis etc. Is there a very basic algorithm for image classification with which I can proceed the implementation ?

The data set I am currently considering is CMU Face Images data set. So aim is to separate images with human faces to those who do not contain human faces. I think I need to learn parameters by supplying positive data set and negative data set as well. I am aware that SIFT, SURF, VJ algorithm etc. are helpful fir extracting features from an image. But I am stuck on how to start i.e. what features are required for face detection, and how to proceed towards implementing them.

  • 2
    $\begingroup$ Would be better if you explained what type of classification you needed. Eg. If all you wanted was bright and dark images then its very different than if you wanted to classify as with faces and without. $\endgroup$
    – av501
    Mar 30, 2013 at 18:00
  • 1
    $\begingroup$ You really need to be more specific about what you are trying to achieve. A basic algorithm is: bool detect_face(img image) { return true; } --- it's easy to implement, but not very useful. $\endgroup$
    – Peter K.
    Mar 31, 2013 at 21:10
  • $\begingroup$ @PeterK. I have updated my question with more details. $\endgroup$
    – krammer
    Apr 1, 2013 at 4:50
  • $\begingroup$ Perhaps you should try learning those features. SVM seems to be a popular way of doing it nowdays. Beware of the time consumption for that though. Do some basic pre-processing and use a background subtraction library to reduce the number of features beforehand. Note: SIFT/SURF/ORB all will give you a large set of features which you should use in order to get your basic set. After that pre-processing in order to reduce number of features is generally necessary. $\endgroup$
    – Naresh
    Apr 1, 2013 at 11:27
  • $\begingroup$ OK, so you are looking for face detection algorithms. I've added that tag to the question. Will dig a bit now for a proper response. $\endgroup$
    – Peter K.
    Apr 1, 2013 at 15:24

1 Answer 1


Have a look at this report by Marques (PDF). In particular, have a read through section 1.4. It describes several different aspects to the face detection problem. You might want to decide what scenario you're after. Fully unconstrained (uncontrolled environment) face detection is hard to do accurately.

As a first approach, try just doing color-based segmentation. See this paper by Singh, Chauhan, Vatsa and Singh for some details about how to use color to segment your images for face detection.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.