Hi, I am the beginer for doing simple image subtration to obtain defect image. Is it possible to make it only show the defect image if nothing difference wont show?

clc

clear

close all

warning off;

x=imread('origin.jpg');

y=imread('capture.jpg');

[g, c, d]=size(x);

y=imresize(y,[g,c]);

subplot(1,3,1);

imshow(x);

title('origin image');

subplot(1,3,2);

imshow(y);

title('capture image');

subplot(1,3,3);

imshow(x-y);

title('defect occur if difference colour shown');

Answer 1

[yanqi liu]

yes，sir

Is it possible to make it only show the defect image if nothing difference wont show?

may be add some judge rule，such as

clc
clear
close all
warning off;
x=imread('origin.jpg');
y=imread('capture.jpg');
[g, c, d]=size(x);
y=imresize(y,[g,c]);
subplot(1,3,1);
imshow(x);
title('origin image');
subplot(1,3,2);
imshow(y);
title('capture image');
bw = find((x-y) > 1);
if numel(bw) > 1
    subplot(1,3,3);
    imshow(x-y);
    title('defect occur if difference colour shown');
end

[CHEE HOON SEOW]

OMGGGG!!!! It's works!!!! Thank you!!!!

[Image Analyst]

Yes, but beware that (x-y) will clip to zero and not go negative.

Answer 2

[Image Analyst]

I suggest you use imabsdiff().

diffImage = imabsdiff(x, y);
threshold = 5; % Images must be at least 5 gray levels different to be considered a defect.
mask = diffImage >= threshold;
imshow(mask)
% Get area of defects
props = regionprops(mask, 'Area')
allAreas = [props.Area]
if ~isempty(props)
    fprintf('Found %d defect areas.\n', length(props))
else
    fprintf('Found no defect areas.\n')
end

[CHEE HOON SEOW]

Hi, I am sorry due to the code may be too deep for me and may I know how to use it, where should I put it?

I am mechanical student and recently works in final year project which slightly touch on the machine vision on detecting defect on the surface conveyor track.

Do you have any recomendation where to learn more about machine vision on defect detecting? Due to the matlab software are new for me and I just know to use a simply image subtraction.

[Image Analyst]

You accepted @yanqi liu's answer so I'm just replacing what was below

clc
clear
close all
warning off;
x=imread('origin.jpg');
y=imread('capture.jpg');

with a more robust version. I'm assuming you could do that because you accepted the answer. My version will handle differences brighter or darker than the background. Because she did (x-y), it will only find pixels where x is brighter than y, not pixels where x is darker than y. So that's why I used imabsdiff(x, y) instead. You said "just know to use a simply image subtraction" -- well, that's just what imabsdiff() does. It simply subtracts, allowing for negative values unlike (x-y), and takes the absolute value. So you get the difference regardless if the defect is brighter or darker than the reference, no-defect image. But I don't want just any difference because there will be image noise and there will be thousands of pixels that are just a gray level or two different than the reference image. So I require the difference to be at least 5 gray levels different before it's considered as a genuine defect. You can adjust that number if you want.

Full code would then be

% yanqui's code first:
clc
clear
close all
warning off;
x=imread('origin.jpg');
y=imread('capture.jpg');

% My changes below:
diffImage = imabsdiff(x, y);
threshold = 5; % Images must be at least 5 gray levels different to be considered a defect.
mask = diffImage >= threshold;
imshow(mask)
% Get area of defects
props = regionprops(mask, 'Area')
allAreas = [props.Area]
if ~isempty(props)
    fprintf('Found %d defect areas.\n', length(props))
else
    fprintf('Found no defect areas.\n')
end

[CHEE HOON SEOW]

it works!!!

at first I am getting some error such as the image obtained need to be convert it into grayscale image only able to perform the code given. THANK YOU

By the way, is it possible to add on the for loop function in the code to make it loop in multiple image? I had been try the simple code but it keep error told that not allow to code like this

please advice

[CHEE HOON SEOW]

clc

clear

close all

warning off;

for k=1:10 %HERE IS I AM TRYING TO ADD FOR LOOP FUNCTION BUT IT FAIL

y=strcat(num2str(k),'.png');

x=imread('reference.png');

x_1=im2gray(x);

y_1=im2gray(y);

% My changes below:

diffImage = imabsdiff(x_1, y_1);

threshold = 18;

mask = diffImage >= threshold;

% Get area of defects

props = regionprops(mask, 'Area');

allAreas = [props.Area];

if ~isempty(props)

fprintf('Found %d defect areas.\n', length(props));

imshow(mask);

else

fprintf('Found no defect areas.\n');

end

end

------------------------------------------------------------------------------------------------------------------------------------

Error using im2gray (line 42)

Invalid argument at position 1. Value must be numeric.

Error in testing (line 10)

y_1=im2gray(y);

[Image Analyst]

Like it says, the input to im2gray(), which is your y, must be an image, a numerical matrix, like of type uint8. You're passing it a character array, like a filename. Why are you doing that?

[CHEE HOON SEOW]

This is the original code written from you

% yanqui's code first:

clc

clear

close all

warning off;

x=imread('origin.jpg');

y=imread('capture.jpg');

% My changes below:

diffImage = imabsdiff(x, y);

threshold = 5; % Images must be at least 5 gray levels different to be considered a defect.

mask = diffImage >= threshold;

imshow(mask)

% Get area of defects

props = regionprops(mask, 'Area')

allAreas = [props.Area]

if ~isempty(props)

fprintf('Found %d defect areas.\n', length(props))

else

fprintf('Found no defect areas.\n')

end

I am not sure why getting error message and search on it told that the image need to be convert it as grayscale, therefore i try it using rgb2gray() function to convert and do the imabsdiff function, it works

now, I am trying to add in the for loop to make the y image become variable and change the image according inside the loop for doing image substraction.

after that i still getting error, the matlab shown error on function rgb2gray() and told that in error message and i try to change it to im2gray() error again.

due to i have 100+ of image for the y so that i am trying to add in the for loop but it turn out errors

[Image Analyst]

You keep forgetting to attach your original images. I'll check back tomorrow.