Prev: shadow removal for traffic sequences
Next: Solve a LMI problem with some variables?(Please help)
From: Rui Xavier on 13 Feb 2010 08:38
http://drop.io./snc38ic#
From: ImageAnalyst on 13 Feb 2010 09:56
On Feb 13, 8:38 am, "Rui Xavier" <ruix...(a)yahoo.com> wrote:
> http://drop.io./snc38ic#

-----------------------------------------------------------------
OK, it looks like you've done some edge detection and found the
centroids of the edge blobs. I'm not sure why you say that you "want
jus the single centroid of the ear image" - not sure how that helps.
Besides, how do you know that the ear is even centered in the field of
view? So the centroid of a particular ear is meaningless until you
get it centered. And you can see from your example that the centroid
does not necessarily lie in the foreground of a blob, for example with
"C"-shaped blobs. Maybe you can crop the edge image by the bounding
box of ALL the blobs and then find the centroid with respect to that.
Of course you can make up a feature vector of all kinds of
measurements. I can't help you with your neural network. But do you
have any specific questions?
From: Rui Xavier on 14 Feb 2010 02:21
hi thanx alot!!!i had assumed that i needed a single centroid of the ear image but with all the inner edges as well...however it isnt possible i guess so will settle for the multiple centroids..am thinkin of mkin my feature vector with distances between centroids rather dan distance of points on t outer ear edge..is it a good an effective method??besides our ear images were given to us and we have cropped them randomnly..so will it become a problem of any sort..could you help with a solution...awaitin a quick reply ...thank u
From: ImageAnalyst on 14 Feb 2010 02:26
On Feb 14, 2:21 am, "Rui Xavier" <ruix...(a)yahoo.com> wrote:
> hi thanx alot!!!i had assumed that i needed a single centroid of the ear image but with all the inner edges as well...however it isnt possible i guess so will settle for the multiple centroids..am thinkin of mkin my feature vector with distances between centroids rather dan distance of points on t outer ear edge..is it a good an effective method??besides our ear images were given to us and we have cropped them randomnly..so will it become a problem of any sort..could you help with a solution...awaitin a quick reply ....thank u

------------------------------------------------------------------------------------------
2 minutes later - how's that for fast?
Well if it works it's good, right? And if it doesn't work it's not
good, right? I've never heard of this application before so you may
be treading somewhat new ground here, just with standard tools. All
that counts is if it works for you in your situation. Nothing else
matters.
From: Rui Xavier on 14 Feb 2010 02:43
dat was unexpectedly fast..yes i am a bit apprehensive about moving along because i want to make sure im doing it right...well if u can help me(which i am deeply grateful to so far)..id like u to check the link http://drop.io./snc38ic# hope u can help me in findin the optimum edge detection code...i am havin a prob with the reflections and the hair..i just need the edges(inner and outer)...
First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4
Prev: shadow removal for traffic sequences
Next: Solve a LMI problem with some variables?(Please help)