5 Replies - 3448 Views - Last Post: 26 November 2007 - 09:50 AM Rate Topic: -----

#1 tommy_chai   User is offline

  • New D.I.C Head

Reputation: 0
  • View blog
  • Posts: 13
  • Joined: 05-November 07

Image processing on Mouth ROI

Posted 20 November 2007 - 06:52 AM

The attachment "m-004-1.zip" contains "ipskin.pgm" and "m-004-1.pgm".
"m-004-1.pgm" is the mouth region after opening(3x3) and closing(2x2) and labeling.
"ipskin.pgm" is the raw image. Both of the image in binary format.
guys, is it possible to do some image processing techniques and obtain only the mouth region at the middle?
I have tried to get the largest connected component but i got the moustache instead of mouth. What terms can i set so that every image will get the lips only and the rest (moustache and small connected component) become black background?

Thank you.

Best regards,
Tommy

Attached File(s)



Is This A Good Question/Topic? 0
  • +

Replies To: Image processing on Mouth ROI

#2 tommy_chai   User is offline

  • New D.I.C Head

Reputation: 0
  • View blog
  • Posts: 13
  • Joined: 05-November 07

Re: Image processing on Mouth ROI

Posted 20 November 2007 - 07:08 AM

Thread has been deleted

This post has been edited by tommy_chai: 20 November 2007 - 07:47 PM

Was This Post Helpful? 0
  • +
  • -

#3 NickDMax   User is offline

  • Can grep dead trees!
  • member icon

Reputation: 2255
  • View blog
  • Posts: 9,245
  • Joined: 18-February 07

Re: Image processing on Mouth ROI

Posted 20 November 2007 - 02:39 PM

neat problem...

The answer is yes... and no... Basically you can use image recognition techniques to look for things that are "lip shaped". This is similar to how OCR works.

The problem with using some kind of bitmap method is the large area of white at the bottom.

thinking outloud...

Writing an OCR like program would take a while... but what about just looking at a ratio of height to width...
Was This Post Helpful? 0
  • +
  • -

#4 tommy_chai   User is offline

  • New D.I.C Head

Reputation: 0
  • View blog
  • Posts: 13
  • Joined: 05-November 07

Re: Image processing on Mouth ROI

Posted 20 November 2007 - 07:50 PM

View PostNickDMax, on 20 Nov, 2007 - 02:39 PM, said:

neat problem...

The answer is yes... and no... Basically you can use image recognition techniques to look for things that are "lip shaped". This is similar to how OCR works.

The problem with using some kind of bitmap method is the large area of white at the bottom.

thinking outloud...

Writing an OCR like program would take a while... but what about just looking at a ratio of height to width...


Image recognition technique, do you mean template matching?
Do you mind to explain what is OCR?

Height to width ratio is one of the way. However, this does not guarantee that we can get the mouth region for every individual, right?

Thanks for your idea. Keeps going.
Was This Post Helpful? 0
  • +
  • -

#5 NickDMax   User is offline

  • Can grep dead trees!
  • member icon

Reputation: 2255
  • View blog
  • Posts: 9,245
  • Joined: 18-February 07

Re: Image processing on Mouth ROI

Posted 20 November 2007 - 09:46 PM

OCR - optical character recognition.

Using the height-width method would not be a very good general solution. but it might work for the example and versions like it.

The more general solution I think would be something akin to OCR. Basically you define a set of acceptable shapes based upon vector templates. You then analyze the image and create vector representation of the regions of the image. Then you compare them with templates giving each a score and then see which has the highest score. I am of course speaking about something I know VERY little about. You will have to research the techniques on your own.
Was This Post Helpful? 0
  • +
  • -

#6 tommy_chai   User is offline

  • New D.I.C Head

Reputation: 0
  • View blog
  • Posts: 13
  • Joined: 05-November 07

Re: Image processing on Mouth ROI

Posted 26 November 2007 - 09:50 AM

View PostNickDMax, on 20 Nov, 2007 - 09:46 PM, said:

OCR - optical character recognition.

Using the height-width method would not be a very good general solution. but it might work for the example and versions like it.

The more general solution I think would be something akin to OCR. Basically you define a set of acceptable shapes based upon vector templates. You then analyze the image and create vector representation of the regions of the image. Then you compare them with templates giving each a score and then see which has the highest score. I am of course speaking about something I know VERY little about. You will have to research the techniques on your own.

Thanks anyway for the explanation. It's something like template matching and i iwll figure itout. Thanks again.
Was This Post Helpful? 0
  • +
  • -

Page 1 of 1