Which Features extraction method will be more accurate for classifying the images based on their Clarity?
I want to classify Images into Good clarity or Bad Clarity (Clarity Check) by using SVM. I have used the HOG feature extraction method, but its accuracy is very poor.which feature extraction method will be suitable for image clarity check? 
can you explain "Clarity Check" ? also, maybe a (small) example image wouldbe helpful
@berak : Clarity check means that the image is classifying into good or bad based on their Clarity. Bad images will be blurred, color faded or shaded images, Noisy images (quality of the image will be poor, so user can't work on detecting ROI from such images).
indeed hog features can't solve this. maybe have a look at algorithms like "BRISQUE" or "Naturalness Image Quality Evaluator(NIQE)"
@berak : I have also used HOH with SVM to classify whether a image contains human or not .The images were captured from roof and some images were captured from side wall ,hence the images didn't contain the full body of the human. I trained some images that contains human and some images without human by using HOG and SVM. (training database =30 images for each)..But the output is poor.. Is this not the right method to classify the task ?
and how is that related to your question ?
and expecting something from 30 images only is outright silly.
you need like thousands (see e.g. INRIA database)