how to get Integrated circuit pads in image?

I want to detect integrated circuit pads in this image. But I don't understand how to get only pads. is there any method or suggestion. thank you

[DGM]

In how many images? Is it just this one? Changing the plating finish and mask color will change the requirements.

Are they all heavily damaged JPGs?

Am I right to assume you're looking for all component pads (chip, QFP, PTH)?

I'm assuming this is just an exercise based on a synthetic image, and not an actual practical application where reliability matters.

[mehmet baki]

just this image. there are 2 ICs in image.

Answer 1

[Image Analyst]

There are hundreds of papers on that. See http://www.visionbib.com/bibliography/applicat837.html#Inspection%20--%20Chips,%20Wafers,%20PCB,%20PWB,%20VLSI,%20IC,%20Disks,%20etc.

Pick one and code it up. Since they're published papers that people spent years on, I'm sure their algorithm would be more robust than anything you'll find here.

If you don't want to do that then I'd try just using the Color Thresholder app (on the Apps tab of the tool ribbon) to get the lighter things. Then I'd use imerode to disconnect the pads from the lines, then use regionprops to check the circularity of the blobs and throw out the non-circular blobs.

See my Image Segmentation Tutorial in my File Exchange for an example of how to filter blobs based on some attributes:

Image Analyst's File Exchange

It's a generic, general purpose demo of how to threshold an image to find blobs, and then measure things about the blobs, and extract certain blobs based on their areas or diameters.

[mehmet baki]

thanks for answering. I am just looking for morphological operations. I need to turn this image black and white then apply some operations such as opening closing to get only pads but I am struggling to get it

[Image Analyst]

Did you try the steps I gave you? Is this not a real, important real world project where you don't need it to be robust and perfect, but it's just some homework assignment?

[mehmet baki]

ı know morphological operations how they work but this problems make me confuse. this is part of assignment I try to understand how to solve.

Answer 2

[DGM]

This is not robust or applicable to other images or types of packages

% read the image
inpict = imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/1702191/image.png');

% upscale so that we have enough resolution
% for the features to survive morphological operations
rescalek = 3;
inpict = imresize(inpict,rescalek);

% try to get a mask from what remains of H
[H,~,~] = rgb2hsv(inpict);
mk = H>0.15 & H<0.20;

% clean up the mask
mk = imfill(mk,'holes'); % get rid of holes to keep them from cutting blobs
mk = imopen(mk,strel('disk',3)); % whittle away the traces
mk = bwareaopen(mk,750); % get rid of specks
mk = bwpropfilt(mk,'eccentricity',[0.9 1]); % try to keep only elongated blobs

% we still have a bunch of junk trace chunks left.
% assuming that the only target objects are QFP
% we can maybe use component-scale geometry
% to distinguish the right groups of blobs
mkthick = imdilate(mk,ones(55)); % turn the component footprints into rings
mkthick = bwpropfilt(mkthick,'eulernumber',[0 0]); % select objects with one hole

% combine the masks
mk = mkthick & mk;

% whenever we're done, we can scale things back to original size
% if that's even necessary
mk = imresize(mk,1/rescalek);

imshow(mk,'border','tight')