Hi,

Usually that would depend on what it is your trying to classify.

These papers may help:

http://www.robots.ox.ac.uk/~vedaldi/assets/pubs/chatfield11devil.pdf

http://arxiv.org/pdf/1405.3531v4.pdf

Like Paul said, it depends. HOG is not invariant to in-plane rotation, which may be ok if you are trying to detect pedestrians, which are typically not upside down. On the other hand, if you are trying to recognize objects form aerial images, invariance to rotation may be crucial. SIFT and SURF are invariant to rotation. SIFT can be a bit more accurate, but SURF is much faster.

Hello!Thank you all.Apart from Hog,SURF and SIFT how can histogram be used for feature extraction?

Hi sheema,

U may try with Local binary patterns, 

the below link gives more insights on LBP

http://www.ee.oulu.fi/mvg/files/pdf/pdf_730.pdf

https://www.google.co.in/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&uact=8&ved=0CCcQFjAB&url=http%3A%2F%2Fwww.researchgate.net%2Fpublication%2F220508716_Local_Binary_Patterns_and_Its_Application_to_Facial_Image_Analysis_A_Survey%2Flinks%2F0fcfd50d0d8f7bc3bc000000&ei=qsldVKinHZWYuQSw6ICYCQ&usg=AFQjCNG0i3rHB4oMBQCNSFuyy_rJM4owmA&bvm=bv.79189006,d.c2E

enjoy

Sheema it histogram can be used for extracting features. I am using histogram of a color image for extracting features in face segmentation. But in case of color images rgb image is first converted into hsv and then features are extracted. While extracting then features you have to use number of bins you needed which then depends on the size of the image and type.

Thank you. What about greyscale images?

I AM ALSO WORKING ON THIS LINE KINDLY SUGGEST WHAT DID U CONCLUDE FOR CLASSIFICATION?

I went for manual feature extraction. But if you want auto then go for SIFT.

I HAVE USED KMEAN AND FCM FOR SEGMENTATION AND NOW M SUPPOSE TO CLASSIFY NORMAL AND ABNORMAL BRAIN MRI AND IF IT IS ABNORMAL THEN BEGNIN OR MEGNIN..ANY SUGGESTION HOW TO PROCEED??

instead of texture features, you can find more from statistical features.