Hi Mohammed, 

This solution seems to be ad-hoc for your topic. 

http://www.mathworks.com/matlabcentral/fileexchange/46674-real-time-face-recognition-and-detection-system

Bye, 

Alessandro

Hi Mohamed,

are you running the Viola-Jones detection in every frame? If yes, consider using the face detection only as an initialization for some sort of (feature-point) tracking algorithm - running a face detector in every single frame is computationally expensive. Tracking only a few features across frames will give you much faster results.

I found an example of Viola-Jones + KLT tracker with MATLAB, you can have a look in the link below. Although it uses the Computer Vision System Toolbox, which you might not have installed, it should give you an idea of the type of algorithm pipeline I explained.

http://de.mathworks.com/help/vision/examples/face-detection-and-tracking-using-the-klt-algorithm.html

Try to implement with V-Jones combination with KLT and apply detection method for a group of frames instead of single frame.

face detection using color information is the fastest approach, then if it is a possible face you may confirm using other more accurate methods such as the Viola-Jones.

and as Michael has said you do not have to check for every frame, do face tracking in the next frames, it is much faster.

Hi

I hope the attached files be useful

Thanks

Hi,

Check this site

http://www.face-rec.org/algorithms/

Regards

Thank you very much everyone.......helped me big time 

If your video stream is acquired in RGB color, you can consider reducing the false negative instances by working on a different color space and limit the search to those regions that match human skin chroma. This was also hinted by Ahmad Hassanat in the above post.

Option 2:

MATLAB is an interpreted language. However, it provides a bridge to low-level languages (C, C++) through its "mex" file generator which can highly likely accelerate the execution of the Viola-Jones algorithm.

Better Algorithm is from my perspective is:

Voila Jones "Integral Image based MB-LBP".

its faster in context of retaining more robust features due to Local and Global feature description.

To speed-up the face detection process:

First you can use motion and color analysis.

Second, for gray scale images, before applying your preferred face detection algorithm, you can use a simplest and faster face detection algorithm, which has an high detection rate and a high false alarm rate. The main face detection algorithm, which has a low false alarm rate is used to verify the proposed face images.

A last and efficient way to speed-up the face detection process is to use a grid on the image. During a first phase only the intersection points of the grid are tested. When the score or probability of a tested point is higher than a threshold a local and exhaustive search is done on the four adjacent squares. For instance if a 3x3 grid is used, the speed-up factor is around 9.

For more details see:

https://www.researchgate.net/publication/3193210_A_Fast_and_Accurate_Face_Detector_Based_on_Neural_Networks

Article A Fast and Accurate Face Detector Based on Neural Networks

@Raphaël Feraud

Thank you for your suggestion. 

is it possible to get the MATLAB implementation of this paper

Would help me a lot 

Thankyou

Unfortunately, there is no MATLAB implementation of this paper.

Regards,

Raphaël

@Raphaël Feraud 

Dear, 

I am having difficulty on implementing your paper, is there any kind of implementation I can use for the moment (any platform),

you should select the feature to work on .