Ask Your Question

Further image processing analysis on roi

asked 2013-09-11 01:27:53 -0500

sp_mic88 gravatar image

updated 2013-09-11 01:36:02 -0500

Moster gravatar image

I have a question to ask: suppose you have a classifier that locate the mouth within the face in an image, this classifier returns multiple ROI as a result. I would like to do further analysis on each of these roi, to determine if, within these, there really is a mouth or not. How could I make this kind of analysis? What are the techniques of image processing to use? I would be grateful if you could recommend some article, or link me the material that may still be useful to me.

edit retag flag offensive close merge delete

2 answers

Sort by ยป oldest newest most voted

answered 2013-09-11 04:44:06 -0500

In the first localize eyes or nose then specify the approximate region of month then localize in this region.

edit flag offensive delete link more


I have already identified the face, and within it the regions of eyes, nose and mouth. For this purpose, I have used cascade classifiers of opencv. The problem is that the mouth classifier returns a lot of false positive. So, my purpose was to find a way to analyze the ROI of mouth, to check if it really contains a mouth. So, for example, I can think to study the color of image, because red pixel are probably pixel of mouth. I can also study the texture of mouth.

sp_mic88 gravatar imagesp_mic88 ( 2013-09-11 10:32:11 -0500 )edit

answered 2013-09-11 01:49:50 -0500

You will need to extract some kind of features (Haar like, hog, edge, or somthing other) and use them to train some classifier (SVM,adaboost). So by that scheme you can make cascade classifier ;)

edit flag offensive delete link more


So do you suggest me to use another classifier? Because I have already trained one, and this returns me that set of roi. My goal was to study these without other classifiers, but analysing with the help of texture or color information (it's only an idea).

sp_mic88 gravatar imagesp_mic88 ( 2013-09-11 02:38:02 -0500 )edit

I have tried to segment the image, but without success... For now, the best approach is the cascade classifier.

Spas Hristov gravatar imageSpas Hristov ( 2013-09-11 02:56:24 -0500 )edit
Login/Signup to Answer

Question Tools


Asked: 2013-09-11 01:27:53 -0500

Seen: 301 times

Last updated: Sep 11 '13