CODEWITHSUNDEEP
classificationface-recognitionfeature-extractionfeature-selection
user1443642
user1443642

Reputation: 31

Feature Extraction for Face Dectection

I'm required to design (extract) 8 features from a collection of face images (see the url below).

http://faculty.ucmerced.edu/mhyang/face-detection-survey_files/face-sample.gif

These images are the results of moving a window (of a fixed size) over a number of original images. Then train a naive bayes classifier using the training set that contains the values of extracted features and class label.

So what features should I extract from the images? Can someone give me some examples?

Upvotes: 0

Views: 174

Answers (2)

Sid
Sid

Reputation: 249

You can extract the basic features such as texture and shape features using inbuilt commands in matlab. More information about this is available in www.mathworks.com/‎

Upvotes: 0

user2683129
user2683129

Reputation: 79

A simple approach would use pixel statistics, namely, the mean and standard deviation (SD), of the raw pixel values (assuming these are greyscale values) of a given region of a face image (e.g. rectangular region containing top or bottom half of a face).

If you extract the mean and SD of four different regions in each image, e.g., two rectangular regions and two circular regions, that gives eight numeric "high level" features.

The mean and SD represent pixel contrast in a region, and are "high level" features because they cover a group/area of pixels (whereas low level features would be to use some/all of the raw pixel values directly as features).

See the following research article for details:

Mengjie Zhang, Urvesh Bhowan, “Program Size and Pixel Statistics in Genetic Programming for Object Detection”. Evolutionary Computation in Image Analysis and Signal Processing. Lecture Notes in Computer Science. Vol. 3005, 2004. pp. 377-386.

Upvotes: 0

Related Questions