You can use this link :

http://spectrum.ieee.org/tech-talk/computing/software/one-million-faces-challenge-even-the-best-facial-recognition-algorithms

Dear Hayder Dibs,Meta-heuristic algorithms are becoming increasingly popular due to their abilities in solving a variety of real world problems. Hybrid meta-heuristic algorithms (such as Genetic algorithm, Particle Swarm Optimization, Ant Colony Optimization, etc) with feature selection can be a good choice for face recognition.

Best Regards.

Soheila Ghambari.

http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=5633844&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D5633844

I am very appreciate from all of you for your answer

There is no best algorithm, please see https://chemicalstatistician.wordpress.com/2014/01/24/machine-learning-lesson-of-the-day-the-no-free-lunch-theorem/.

I totally agree with Maher Ibrahim as what do you mean by "best" algorithm?

Best for highest accurate detection or best for computational time or easy-implemented ??

it is really up to the imagery you will use as an input data however there are some well-recognized algorithms such as svm or neural-net based algorithms (deep learning or active learning based). 

I suggest you to check python libraries for this purpose and if such algorithms do not work well you can code your original algorihtm. 

scikit-image is a collection of algorithms for image processing. It is available free of charge and free of restriction.

Dear Maher thanks for your answers.However, your link very faraway for what I am asking. We have many algorithms recognize it as very affected for face recognition, such as RI-SIFT, SAM, SURF specially RI_SIFT that I created it, it is fantastic for body and face. However, I am looking for a new  other algorithms that have been invited by other research with different parameters> thanks again

my regards 

Dear Mustafa Üstüner , I am thankful for your reply. It is known that the best algorithm should be " Best for highest accurate detection and  computational time:, and the python or its libraries will be not useful if I have no algorithm to try. Dear I tried many algorithms in face and body recognition but I need to try another algorithms. I already invent a new one called RI-SIFT and it is very good to do so with different parameters such as illumination, capturing time, view direction ...etc . but I need to discover what is in researchers mind ideas and suggestion to develop another one

thanks a lot

my regards

Dear Hayder,

This best is your best and let me explain as below.

Just suppose i am working in planning company and 70% accuracy is enough for my research/project .Method A is quite fast and do not need parameters to apply and it achieves 70% accuracy while method B need four parameters to run (to optimize) and accuracy is %90. For such cases as 70% is enough for me, A is best for me. Hope you understand.

Second issue is nobody can say or offer you the "BEST" algorithm without knowing or meeting your data. What kind of features you use as an input? 

I again suggest you to check OpenCV  or if you wish to learn ""NEW" algorithms that may not be best one, you can check this journal http://www.journals.elsevier.com/pattern-recognition

Hope it helps

Dear Mustafa

yes, each one look to what is the best regarding to his work. But in general all scientists and researchers have the same concepts with the word " BEST" and for that you see 70% is enough to your work, I really do not know how it is work.. However, in most of all digital image processing and analysis researches and applications will be not. even if we get 80% for examples applications in LU/LC classifications, prediction of landslides and floods, Urban analysis, geodetic and space researches  or  classification, aerial photography, geometric correction, and so one.It will not be enough. Simply, because you will loss around 30% of the areas,features  of your imagery, i Do nt know what is your application and how 70% is enough, but this your best, in other side my side as I mentioned to you Dear Mustafa is a face recognition, the matter here sensitive and different, here any mistake will find the wrong person. it is clear to any one what is the meaning of the best that am looking for face recognition not like LU/LC can deal with 65-85%,

So, we have here a good website to discuss between us as researchers to add to each other a "NEW IDEAS" and find the "SCIENTIFIC GAPS" and help each other to fill it, not to argue without scientific evidence.

As you can see and read I had to published patents, thats mean all the time I am looking for the "" NEW"" things to "LEARN" this my point. Thank you so much for you and please if you have any thing new in this field (Mobile  face recognition with differences in sensors, capturing time, angle of sensor-object-illumination,sensor' s altitude and sensors attitude) I will be very appreciate if you share it with me.

Hopping to learn something new to share the scientific researches and add to each other new idea and knowledge

Thanks 

Hayder

Dear Hayder

I did not give any details about application however you directly get to "image classification :)  "But i typed "Just suppose" :) 

http://dictionary.cambridge.org/dictionary/english/suppose

anyway, please let us know when you find "the best" algorithm! 

Cheers

Dear Mustafa you can read our Patent in feature extraction under the title " A METHOD FOR EXTRACTING CONTROL POINTS FROM IMAGES" in my publication side of research gate, it is also one of the algorithms lead to improve the face recognition, till I patent a new one....... any question you have, you are most welcome.

thank you so much 

Finally, I would like to thank every one reply to my question.

My regards