Pixel Scan
in [Matlab]
|
Prev: RBF neural network - How to estimate "spread"
Next: Plotting periodical measurement data with jitter to matlab
From: guccikitty on 10 Nov 2009 11:30 we are selling gucci shoes, gucci bags and ugg boots. welcome to join us. http://www.gucci-shoes-bags.com http://www.uggbootshoes.com
From: wilson on 12 Nov 2009 12:36 ImageAnalyst if i change to other picture what should i change in the coding? Because the target seem to run to other place its wont exactly get the target u want? So 1st izzit i need to fix some distance to take the picture so that everytime whenever i scan the face it will get the target exactly the eyes i wan? Is that concept correct?
From: ImageAnalyst on 12 Nov 2009 13:00 On Nov 12, 12:36 pm, "wilson " <wilson_silver3...(a)hotmail.com> wrote: > ImageAnalyst if i change to other picture what should i change in the coding? > Because the target seem to run to other place its wont exactly get the target u want? > So 1st izzit i need to fix some distance to take the picture so that everytime whenever i scan the face it will get the target exactly the eyes i wan? > Is that concept correct? -------------------------------------------------------------------------------------- You may have to tweak the algorithm. It may only work in some set of situations. For example a certain magnification and pose. For example if someone has one eye open, one eye closed, and their mouth open, that algorithm may think the open mouth is the other eye. Or if the person is so far away that the eyes are only a few pixels then it may have problems. Or if their eyes are closed and the camera is looking up their nostrils then it will pick the nostrils as the eyes. What if you have a picture of a yellow smiley face? Do you want to define its "eyes" as real eyes, or reject them because they're not from a real live human? It can get pretty tricky. That's why I said it's not robust. The better able you are to control your subject, the easier it will be.
From: wilson on 12 Nov 2009 21:36 For my research is just concentrate on eyes, so i just can ignore the mouth part! I just need to take the eyes out from the face! The eyes for everyone is the same if i fix a distance example 1m within the camera then i can directly target the eye just like the method u teach me! But i gonna to make sure the eyes are both open in order to get the target. Because if i use manual cropping method it make my project become inapplicable so i need to think of something that can automatically crop the eye out once the target (eyes) match with the original image. Is this method applicable for my project?
From: ImageAnalyst on 12 Nov 2009 22:07
On Nov 12, 9:36 pm, "wilson " <wilson_silver3...(a)hotmail.com> wrote: > For my research is just concentrate on eyes, so i just can ignore the mouth part! I just need to take the eyes out from the face! The eyes for everyone is the same if i fix a distance example 1m within the camera then i can directly target the eye just like the method u teach me! But i gonna to make sure the eyes are both open in order to get the target. Because if i use manual cropping method it make my project become inapplicable so i need to think of something that can automatically crop the eye out once the target (eyes) match with the original image. Is this method applicable for my project? ------------------------------------------------------------------- The method I sketched out should work reasonably well if all your pictures are similar to the test photo you provided. That is, a face taking up most of the picture, eyes both open, mouth closed, etc. |