Prev: how to use textscan in a for loop with header lines increasing
Next: which way is better ?
From: Michael Coughlin on 6 May 2010 22:15
Hi,

There is a lot of information and code on here that has to do with image processing and the like, but I was hoping that someone might point me in a clear direction.

http://drop.io/arcodme

I have posted here 2 images, produced in Matlab using imagesc. I would like to be able to pull from these information such as where white lines occur or more importantly, where clusters occur (like any splotches of white or red). I have the data in raw form, as I produce the plots, but I thought it might be easier to see for all of you in this way, but I am just pointing out that I have a matrix corresponding to the intensity of each point, so that is not a problem.

I hope this was not too vague and can answer any questions you might have.

Michael
From: ImageAnalyst on 6 May 2010 22:47
On May 6, 10:15 pm, "Michael Coughlin" <cough...(a)carleton.edu> wrote:
> Hi,
>
> There is a lot of information and code on here that has to do with image processing and the like, but I was hoping that someone might point me in a clear direction.
>
> http://drop.io/arcodme
>
> I have posted here 2 images, produced in Matlab using imagesc. I would like to be able to pull from these information such as where white lines occur or more importantly, where clusters occur (like any splotches of white or red). I have the data in raw form, as I produce the plots, but I thought it might be easier to see for all of you in this way, but I am just pointing out that I have a matrix corresponding to the intensity of each point, so that is not a problem.
>
> I hope this was not too vague and can answer any questions you might have..
>
> Michael
-----------------------------------------------------------------------------
Can you also post the original image - the monochrome/grayscale
version?

Can you just do the usual image analysis process:
thresholding
bwlabel
regionprops?
It depends on how small a blob you want. Are you willing to accept
single-pixel sized regions?
From: Michael Coughlin on 8 May 2010 14:02
Hi,

>> [L, num] = bwlabel(x.ftmapF,8);
>> RGB = label2rgb(L);
>> imwrite(RGB,'test.jpeg','jpg')

This produced a plot of basically 6 rectangular regions. Do you have suggestions as to how to give bwlabel better options to improve what it finds? Or will delving into regionprops be necessary? As regionprops has a lot of options with it, do you have any recommendations as to which piece of regionprops to try first? I am seeking regions larger than one pixel, as to your question above.

Michael
From: ImageAnalyst on 8 May 2010 15:37
Michael:
Can you also post the original image - the monochrome/grayscale
version? Otherwise it's difficult to threshold or do anything else
(e.g. imregionalmax) that will give you meaningful results if all we
have is the pseudocolored image.

You can use bwareaopen to filter the binary image based on size.

And I don't know if you'll need regionprop(). You haven't really said
if you want to measure anything. If you simply want to visualize
where clusters are then producing a binary map of their locations
would be enough I guess. If you want the coordinates, then either use
find() on the binary image, or regionprops() on the labeled image if
you want the coordinates organized by blob number.

Maybe my script demo on blob analysis will help you understand the
role of thresholding, labeling, and regionprops.
http://www.mathworks.com/matlabcentral/fileexchange/25157

Write back with the URL of your grayscale image if you want more help.

-ImageAnalyst


From: Michael Coughlin on 8 May 2010 15:50
Hi,

Can I ask a really silly question: how?

imshow(array,[],'Xdata',xvals,'Ydata',yvals, ...
'InitialMagnification','fit')
colormap(hot), h=colorbar;
daspect([1 1 1]), axis square;

That is basically what I use to produce these. Is there an easy way to make it monochromatic?

Thank you,
Michael
 | 
Pages: 1
Prev: how to use textscan in a for loop with header lines increasing
Next: which way is better ?