Prev: how can you restart CUMSUM
Next: how can i compute probability map p-man
From: jrenfree on 23 Nov 2009 15:05
On Nov 23, 11:50 am, omo_ewa <morenike.aj...(a)gmail.com> wrote:
> On Nov 23, 2:48 pm, jrenfree <jrenf...(a)gmail.com> wrote:
>
>
>
> > On Nov 23, 11:41 am, omo_ewa <morenike.aj...(a)gmail.com> wrote:
>
> > > On Nov 23, 2:27 pm, jrenfree <jrenf...(a)gmail.com> wrote:
>
> > > > On Nov 23, 10:53 am, omo_ewa <morenike.aj...(a)gmail.com> wrote:
>
> > > > > Hello all,
>
> > > > > I am new to using matlab for any serious coding. However, I will like
> > > > > to be able to recognize specific colors (green, red, blue, black,
> > > > > etc.) in a scene of differentcolorblocks of different colors. So, a
> > > > > sample will be to recognize a yellow-triangle block in a scene of
> > > > > yellow-triangular block, blue-square block, red-circular block. Thank
> > > > > you for your help!
>
> > > > > omo_ewa
>
> > > > Generally, colors are described by a combination of their RGB (red,
> > > > green, and blue) values.  These values range between 0 and 255.  In
> > > > Matlab, an image will typically be given by an MxNx3 matrix that gives
> > > > the RGB values for each pixel of an image.  You will need to determine
> > > > the RGB ranges that coincide with the colors you want to detect and
> > > > use some sort of thresholding in your image matrix to pick those
> > > > colors out.
>
> > > > There are other ways to do it, such as converting the image to HSV
> > > > (hue, saturation, and value) values instead of RGB.
>
> > > Hello jrenfree,
>
> > > Thank you for your help. It will be useful to me later on. But, do you
> > > have a sample that shows even how to start this. I know only so far
> > > how to read an image in to matlab. How do i code something that now
> > > checks for green in an image? What i am saying is that I don't know
> > > what to write. I have been checking online for a sample code i can
> > > work with, but i have not found any yet. I believe my problem is more
> > > basic.
>
> > > Thank you,
>
> > > omo_ewa
>
> > Ok, so let's say you read in an image:
>
> > A = imread('test.jpg');
>
> > The variable A should be some MxNx3 matrix.  Now if we want to pick
> > out all the green values from that matrix, we need to know the RGB
> > values for green.  Solid green would contain no red (R=0), full green
> > (G=255), and no blue (B=0).  We could pick those out by:
>
> > idx = A(:,:,1) == 0 & A(:,:,2) == 255 & A(:,:,3) == 0;
>
> > This is essentially saying to look for indices where the red color
> > values equal 0, the green color values equal 255, and the blue color
> > values equal 0.  The result, idx, is a logical matrix that indexes
> > into A to pick out only the pixels that are green.
>
> Thank you very much, I'll start from there.
>
> omo_ewa

I should add that if you wanted to look at just those pixels, you
could do:

image(A(idx))
From: ImageAnalyst on 23 Nov 2009 15:41
You could use the known RB values and do some thresholding and ANDing
(like jrenfree showed you) - pretty simple but not too robust. See my
demo:
http://www.mathworks.com/matlabcentral/fileexchange/25157
and just adapt it to three color planes instead of one.

Another method is to use k-means clustering like in this MATLAB demo:
http://www.mathworks.com/products/demos/image/color_seg_k/ipexhistology.html

Like jrenfree said, you really need to say what you mean by a certain
color because there are a possible 16.7 million of them in a 24 bit
rgb image and of course we don't have 16.7 million names. So each
name, such as "blue" must cover a RANGE of colors and you need to
define this. It also depends on the image. If you take a picture of
a very simple image it might be clear what blue, red, and black are,
and everybody agrees on them, but if it's more complicated, say a Van
Gogh Painting, it's not so clear what the color names mean because
there is a CONTINUUM of colors, not a small, discrete set of colors.
First  |  Prev  | 
Pages: 1 2
Prev: how can you restart CUMSUM
Next: how can i compute probability map p-man