How do I get the area of a zone covered by disconnected blobs in an image?

I'm trying to get the following dimensions from a specific image of a chip:

• the perimeter of the entire chip
• the area covered by the pins (it's not the sum of every pin's area)
• the area covered by the transistors (the internal black square)

I've found a solution for first point but I don't have any idea for the last two points.

The image is the following:

[Image Analyst]

It depends on how accurate you want it, and what your definitions of the areas are. Does the chip include the black board or plastic it's on? If so you'll need to use better lighting to get rid of the shadows around it. You should also use more uniform lighting or else take a background shot to divide out and correct for the horrendous lens shading you have.

If the area of the pins is not the sum of the area of the pins, then exactly what is it? You said what it's not, but not what it is.

What is the area of the transistors? The white blobs in the middle square? Their bounding box? Or the outer area of the middle dark square?

[mirko]

I explained all in my question, please read it better. The guy who sent me the code below understanded everything and gave me a hint to solve my problem. I also completed the job with my horrendous image so I don't even need help anymore. Thanks.

[Image Analyst]

I've read it several times and it's still ambiguous. Anyway I'm glad that @Sulaymon Eshkabilov made the correct assumptions, though he does compute pin area as the sum of the individual pin areas which you explicitly said was not correct. Since you say it does everything exactly as you want, go ahead and accept his answer.

I'll try to remember not to help you anymore. In fact I went back and deleted my answer I posted a few minutes ago to your latest question about spatial calibration. Good luck though.

By the way, his answer is actually totally wrong. There are several errors in it that give wrong pixel counts. Did you even do a sanity check on it? How can the pins area be nearly the same area as the entire image? I made some modifications to @Sulaymon Eshkabilov code to show you the blobs that he identified as pins and transistors and here is the result (full code is attached.)

Look at all the white blobs in the left image. That's is what he's calling pins. But perhaps you were able to fix it.

[mirko]

It was a wrong code but he gave me a hint that helped me instead of criticizing my question. Could you please repost your answer on my latest question and help me?

Answer 1

[Sulaymon Eshkabilov]

It will be somewhat like this one:

close all;
clearvars
% Image Import:
I = imread('Chip_image.png');
% Grayscale Conversion:
G_Img = im2gray(I);
figure
imagesc(G_Img)
title('Imported Gray Scaled Image')

% Find a Thresh of the chip area from background
B_Img = G_Img > 90;  % Adjust this threshold

% Morphological operations to clean up the binary image:
CB_Img = imclose(B_Img, strel('disk', 5));

% Connected Component Analysis:
CCA = bwconncomp(CB_Img);
stats = regionprops(CCA, 'Area', 'BoundingBox');

% Filter connected components based on area. Still, some adjustment for
% thresholds required
Pin_Threshold = 500;
Transistor_Threshold = 200;

% Initialize variables:
Total_Chip_Area = sum([stats.Area]);
Pin_Area = 0;
Transistor_Area = 0;

% Iterate through connected components
for i = 1:length(stats)
    % Check the area and categorize into pins or transistors
    if stats(i).Area > Pin_Threshold
        Pin_Area = Pin_Area + stats(i).Area;
    elseif stats(i).Area > Transistor_Threshold
        Transistor_Area = Transistor_Area + stats(i).Area;
    end
end

% Display the calc. results:
fprintf('Total Chip Area:  %d \n', Total_Chip_Area);

Total Chip Area:  2013878 

fprintf('Area covered by Pins: %d \n', Pin_Area);

Area covered by Pins: 2006139 

fprintf('Area covered by Transistors:  %d  \n', Transistor_Area);

Area covered by Transistors:  2743  

% Display the segmented chip area
figure;
imshow(CB_Img);
title('Segmented Chip Area');