Prev: color names
Next: No one?
From: ImageAnalyst on 8 Mar 2010 18:08
On Mar 8, 3:58 pm, "Valentine Oradiegwu" <obioradie...(a)yahoo.com>
wrote:
> ImageAnalyst <imageanal...(a)mailinator.com> wrote in message <febd7d93-e599-4c19-bac1-9f9b44395...(a)b7g2000yqd.googlegroups.com>...
>
> ImageAnalyst,
> In school, we are required to write a tracking algorithm that associates blobs from subsequent frames using their average color and centroids! That's why i'm trying to compute the average color after having computed the centroid of the blobs. And no, I have not done some scene calibration. What merits can i enjoy from screen calibration and using CIE values instead of RGB?
>
> I can extract each color channel quite comfortably but does regionprops accept arguments that are not Binary, Labelled, or Connected Component? Your suggestion seems a lot more easier (once i can understand a little bit more) than what i did today. I used the property PixelList to return the coordinates of each pixel on the blobs and reckoned that these coordinates will be same on the RGB image and used these coordinates to extract each color channel, sum them up and divide by number of pixels.
>
> Kindly assist in clarifying your excellent suggestion for the average color computation.
> Many thanks.

-------------------------------------------------------------------------------------------------------------------
Valentine:
This is just straightforward application of standard function calls.
Run my demo on blob detection
http://www.mathworks.com/matlabcentral/fileexchange/25157
and color detection
http://www.mathworks.com/matlabcentral/fileexchange/26420-simplecolordetection
and come back if you have any difficulty in applying what you learned
in there. You'd have to post your images also.
Plus, you can get the mean directly and don't have to worry about
PixelList (unless your version of MATLAB hasn't been updated in the
last two and a half years, in which case it's time to do it.)

regionprops only takes a labeled image and an optional original
monochrome image. What else do you think it *should* take?
Good luck,
ImageAnalyst
From: Valentine Oradiegwu on 11 Mar 2010 03:30
Dear Image Analyst,
Many thanks. The problem has been solved successfully! In a tracking application written in matlab, I have successfully matched features of objects in previous frame to objects in current frame. Which function(s) could be used to propagate a consistent label from blobs to blobs that match certain features?
Many thanks!
From: ImageAnalyst on 11 Mar 2010 06:17
Valentine Oradiegwu
I'm not really in the field of tracking so I'm not sure. I believe
(I've heard) that they make up 3D volumes of the binary images with
one axis being time and then they do labeling in 3D, assuming that the
object didn't move so much that it doesn't overlap it prior position.
If the objects do move a lot, there are sophisticated projection/
prediction algorithms to try to guess which blob was where in the
prior slice of the 3D image (i.e. at a different time). Understand?
That's about all I know - I don't know the details.

You can find virtually every article written on the subject here:
http://iris.usc.edu/Vision-Notes/bibliography/contents.html
Look especially at sections 16, 17, & 18.
-ImageAnalyst
From: Valentine Oradiegwu on 11 Mar 2010 13:38
Hello ImageAnalyst,
Many thanks.
Cheers!
From: stephanie scicluna on 26 Mar 2010 07:25
"Valentine Oradiegwu" <obioradiegwu(a)yahoo.com> wrote in message <hnbdat$54k$1(a)fred.mathworks.com>...
> Hello ImageAnalyst,
> Many thanks.
> Cheers!

Hi,

May i ask how did you manage to object the average colour?

I have something similar. this is how i carried it out -> i have an image which i have converted to binary and added some touch ups to fill in the blanks etc.
Now i have labelled my objects, placed a bounding box around and obtained these values.
I have found the rgb values for the whole image and found the average values but when it comes to just that object i am stuck.

could someone help out pls?

i tried using p = objectProperties(k).PixelIdxList; but get a list of numbers which are quite high like 44778, 44779, etc.. and have no idea what these are!
First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4
Prev: color names
Next: No one?