4

I'm curious about finding the eyes from an image. Let's say I have an thresholded image and I have 3 blobs representing two eyes and the mouth. Is there a way to detect which of the region of pixels is eye region and approximate it's center?

onetwo12
  • 2,140
  • 5
  • 21
  • 32
Madalin
  • 355
  • 5
  • 14

4 Answers4

2

I am not sure if I follow your question right! Do you want to process a gray level/ color image or a binary image.

You might find the following links relevant if you are trying to locate fiducial points on face images

1) http://www.learnopencv.com/facial-landmark-detection/

2) http://danielnouri.org/notes/2014/12/17/using-convolutional-neural-nets-to-detect-facial-keypoints-tutorial/

3)http://cmp.felk.cvut.cz/~uricamic/flandmark/

eyebies
  • 51
  • 3
2

I would actually recommend you to use dlib library.

I used it recently. You can use it very easily for object detection and if you want to train it for your own objects it is fairly easy using imglab.

http://blog.dlib.net/2014/02/dlib-186-released-make-your-own-object.html

See this.

Anant Simran Singh
  • 561
  • 4
  • 12
2

You can check the following blog which is much equipped with facial landmarks detection including eye regions. It will be very much easier to find approximate center of eye. It gives six points (37-42 left eye & 43-48 right eye) of eye along the eye lid which can be used to calculate the eye center.

enter image description here

Detect eyes, nose, lips, and jaw with dlib, OpenCV, and Python

There are much more you may find them very much interesting and helpful.

Ashfak Md Shibli
  • 175
  • 1
  • 7
1

OpenCV is the acronym of Open Computer Vision . this is a project that works on face-recognition pattern-recognition face-traking and much much more.

There is a well-written documentation and a lot of examples all over the web. I suggest you to take a look at this.

It works with Java,c++ and python.

Search for eye-traking if this is your main interest.

Gabrio
  • 238
  • 1
  • 3
  • 15
  • With some image filtering and thresholding i managed to isolate blobs containing eyes, but i don't know how to verify if that blobs are more likely to be a mouth blob or an eye blob, i'd like to test that analyzing blob shape – Madalin Jun 18 '16 at 06:04
  • 1
    [this example](http://docs.opencv.org/2.4/doc/tutorials/objdetect/cascade_classifier/cascade_classifier.html) shows how to find , thanks to cascada classifier, an array of eyes associated to their position. the classifier is a way to use a 'machine learning' algorithm in order to find something like an eyes in a frame. – Gabrio Jun 18 '16 at 08:19