How do I find the pixel size of connected components in an image and then compare pixel size of each connected component with the pixel size of connected components of the next frame and choose the approximately similar size connected components?

So,I am reading a video and then saving its frames and in each frame I am finding the connected components and then labelling them.Now lets say I have the image below.In this the pixel size of green cloud on left will be compared with pixel size of both green and red cloud in the second image.If the pixel size of red cloud in second image matches with the pixel size of green cloud in first image then they should save as a pair.Similar procedure needs to be followed for the red cloud in first frame.I have saved these frames but I don't know how to find the pixel size of these green and red clouds separately and then compare them with the next frame's clouds.How can I do that?

Answer 1

[Image Analyst]

It looks like you've already segmented the image into regions and then have labeled the image with bwconncomp() or bwlabel() already, and then are displaying with a color map or perhaps created an RGB image with

% Let's assign each blob a different color to visually show the user the distinct blobs.
coloredLabels = label2rgb (labeledImage, 'hsv', 'k', 'shuffle'); % pseudo random color labels

So, just put your labeled image into regionprops(), and get the areas

props = regionprops(labeledImage, 'Area');
allAreas = [props.Area];

Now that is just the areas. You don't know what area it corresponds to because as the blobs move around, what was blob #1 in one frame might suddenly become blob #2 or #9 or some other ID number. This is because of how the blobs are found. So you're going to have to do tracking if your blobs move around. You can do this if you also find the centroids and if the blob doesn't move very much from one frame to the next so you can just assign the blob numbers to the closest blob from the prior frame.

[Prachi Sharma]

This is what I have done till now.From here I need to find the pixel size and then compare them.

%%compute connected components

CC=bwconncomp(bw);

lab=labelmatrix(CC);

s=regionprops(lab,'centroid');

centroids = cat(1, s.Centroid); figure imagesc(lab)

[Image Analyst]

Did you see my code? Try this:

CC = bwconncomp(bw);
props = regionprops(CC, 'Centroid', 'Area');
allAreas = [props.Area];
allCentroids = [props.Centroid];
xCentroids = allCentroids(1:2:end);
yCentroids = allCentroids(2:2:end);

Now for that frame, you have a bunch of centroids - one for each blob. Assuming the number of blobs and the blob labels don't change, you can simply save the centroids in an array

xCentroidsByFrame(frameNumber, :) = xCentroids;

This is not very robust because it requires you to know in advance the number of blobs, and requires that they don't move around so much that their label changes. However if you don't like these restrictions then you're going to have more complicated code that handles tracking, and that's too involved for me to supply/write that code for you. Good luck though.

[Prachi Sharma]

Yeah,I need to do the tracking as this is all about the tracking of moving clouds.Thanks.

[Prachi Sharma]

Hey you said that this will work if the blobs are the same and they don't change much so it means we can calculate pixel size of these blobs using area,just like you did for centroid?

[Image Analyst]

Yes, and you can ask regionprops for both the area and centroid at the same time

props = regionprops(labeledImage, 'Area', 'Centroid');

[Prachi Sharma]

Right now I tried to do this and when I make it to read a video and do watershed segmentation on it and then perform the following code as given below.It gives an error.Can you tell me why is it giving this error.I have first mentioned the code and then the error-

%compute connected components

CC=bwconncomp(bw);

lab=labelmatrix(CC);

%% FINDING CENTROID OF CONNECTED COMPONENTS(need to work on lab)

s=regionprops(lab,'centroid');

centroids = cat(1, s.Centroid);

figure

imagesc(lab)

hold on

plot(centroids(:,1),centroids(:,2), 'b*')

hold off

%% %% props = regionprops(CC, 'Centroid', 'Area');

allAreas = [props.Area];

allCentroids = [props.Centroid];

xCentroids = allCentroids(1:2:end);

yCentroids = allCentroids(2:2:end);

xCentroidsByFrame(i, :)=xCentroids;

end

Error is- Index exceeds matrix dimensions.

Error in tryy (line 119)

plot(centroids(:,1),centroids(:,2), 'b*')

It ran for one video only,for other video it starts giving this error

[Prachi Sharma]

Also I tried running your code alone after finding the connected components that is alone on CC(removing my part of code of finding centroid of connected components,it doesn't save anything then.

[Image Analyst]

Use the debugger. How many columns are in xCentroidsByFrame, and how many elements are in xCentroids? I bet they don't match. You will almost certainly have a different number of centroids in each frame so that won't work. Perhaps xCentroidsByFrame should be a cell array instead. http://matlab.wikia.com/wiki/FAQ#What_is_a_cell_array.3F

[Prachi Sharma]

Actually the centroids is not saving anything.Initially it worked for one video and now it is just not working and not saving anything in centroids.

[Prachi Sharma]

Hey I ran your code for centroid and area.The code is given below.It is giving two values for centroids as there are two clouds but it is giving only one value for area.Since there are two clouds it should be giving two areas but it is not.Can you tell me what did I miss here?I have also attached pic of the values please see them.

<<

>>

props = regionprops(lab, 'Centroid', 'Area');

allAreas = [props.Area];

allCentroids = [props.Centroid];

xCentroids = allCentroids(1:2:end);

yCentroids = allCentroids(2:2:end);

xCentroidsByFrame(i, :) = xCentroids;

xAreas=allAreas(1:2:end);

yAreas=allAreas(2:2:end);

xAreasByFrame(i,:)=xAreas;

[Prachi Sharma]

??

[Image Analyst]

Attach your lab array in a .mat file and I'll see what I can do later today with it to figure it out.

[Prachi Sharma]

Below attached is my whole code,you can run it for any cloud video.Also I have attached lab and CC matrix.I need to find the centroid and area of blobs in every frame and save them for further comparison in future code.