How to count dots and mark those dots with crosses in an Image?

I have an image with bright dots over a dark back ground, I have created a ROI in it and then created a binary image, Now I want to count number of dots in this image inside ROI region and produce a result like "desired_result.jpg", any ideas or suggestions how can I do it?

Answer 1

[DGM]

As far as I can tell, your sample image is corrupted. Either way, you can do this:

% find centers of objects
S = regionprops(mybinaryimage,'centroid');
C = vertcat(S.Centroid);	
numspots = numel(S) % count objects

% show image and plot markers
imshow(mybinaryimage); hold on
plot(C(:,1),C(:,2),'x')

[Mohsin Zubair]

this is sample image.

[DGM]

It's still the same thing, really. Just apply the operation to the area selected by the ROI.

% i just painted the roi lines out of the screenshot
mybinaryimage = rgb2gray(imread('binary_image.png'))>128;

imshow(mybinaryimage); hold on % show image
R = drawpolygon(); % draw ROI
mybinaryimage = mybinaryimage & createMask(R); % mask off unselected area
delete(R) % remove ROI from figure

% find centers of objects
S = regionprops(mybinaryimage,'centroid');
C = vertcat(S.Centroid);	
numspots = numel(S) % count objects

% show image and plot markers
plot(C(:,1),C(:,2),'x')

Of course, if you only want to draw the selected dots, just re-show the new copy of mybinaryimage before plotting the markers.

Either way, the markers and counts are the same

[Mohsin Zubair]

Looks great, I will test this out.

[Mohsin Zubair]

Is there any way to recreate ROI automatically from the image instead of drawing rui manually like we did in code?

[DGM]

Maybe, but I'm not sure of an ideal way. If it needs to work on a range of images, then you'd need to consider more than one in order to develop a method that works generally. If it only needs to work on this one image, then I don't see the point.

[Image Analyst]

@Mohsin Zubair, we don't know because you didn't post the grayscale or RGB image over which you manually drew the ROI, like perhaps with imfreehand() or drawpolygon() or similar. Can you post that and tell us in words the strategy you used in drawing the outline of the ROI? It looks like some of the dots were encircled, and some were not, and we have no idea why it is like that unless we see the original, non-binarized image.

[Mohsin Zubair]

@Image Analyst I used drawfreehand() to draw these ROI and then I used export_fig() an external function to export the orignal image along with ROI on it as a tif image, Orignal image is also in tif formate, There are lot of these images and now I want to do dot analysis on them like DGM told the method but now the ROI is embeded permanently in the pixels. Can you suggest a way to either recreate ROI automatically or perhaps even more better, save Image and ROI as an overlay in tif formate instead of permanentaly embedding ROI in pixels of image?

[DGM]

Why do you need to save the ROI? Are all of the images aligned/registered such that they can use the same ROI?

If you just want the mask to save separately, you can just use

thisisthemask = createMask(R);

[Mohsin Zubair]

Well I just need an image file to keep visual record of the ROI that was made, and all images are different with different ROI in them drawn using drawfreehand().

[Mohsin Zubair]

Also another purpose is that if these images are to be sent to someone else they should have previous ROI information and they should be able to use that ROI which was previously drawn and then do furthur analysis.

[DGM]

As far as saving the ROI in an associated image, that could still be done using createMask() if you wanted. Just save the mask using a non-destructive image format. If you're using TIF, that would be perfectly fine.

As far as trying to automate ROI generation, that's going to be a problem without knowing what the other images look like or understanding the object/signal from which the images are derived. To be succinct, in order to make something that best locates a region of interest, one would need to know conceptually what the interest is.

[Image Analyst]

@Mohsin Zubair, so it looks like your original images are pretty contrasty black and white, though some of the blobs might be brighter or fainter than the others. But again, you have given us no idea of how you decided to outline those particular blobs and not the others. Do you just want to take a large grouping that is roughly in the center of the image? Or can the ROI be anywhere in the image? Is there more thought put into deciding which ones to include than just taking the blobs within a radius of half the image from the center? If so, what is your thought process in deciding which ones to include and which to exclude? We need to know that if we're to develop any sort of automatic way of getting the ROI.

Also, include the original image because the screenshot with the blue stuff in the overlay, while good for showing us which blobs you chose to include, won't be any good when we try to develop an automated method.

[Mohsin Zubair]

@Image Analyst I solved ROI problem by just saved ROI information externally in and then load this information later and draw ROI from this saved ROI information as it doesn't seem feasible to draw ROI automatically any other way, As for how do I determine the DOTS I draw ROI where I need and then do analysis on them