To Mark Matched Blue Pixels in 2 Images

Hello,

I would like to mark matched blue pixels in 2 images.

The images are attached. Here, what I want to have is to mark matched points which are blue. (I am not sure if the blue is the correct approach because the colors are not the same)

As a result , I would like to see it as below with the red marks on the matched pixels;

I need to use rgb2hsv convertion but it didn't work.

My current code is:

startingFolder = 'C:\Users\Development\Desktop\dneme';
if ~exist(startingFolder, 'dir')
  % If that folder doesn't exist, just start in the current folder.
  startingFolder = pwd;
end
defaultFileName = fullfile(startingFolder, '*.jpg');
[baseFileName, folder] = uigetfile(defaultFileName, 'Select a file');
if baseFileName == 0
  % User clicked the Cancel button.
  return;
end
fullFileName = fullfile(folder, baseFileName)
original = imread(fullFileName);

startingFolder = 'C:\Users\Development\Desktop\dneme';
if ~exist(startingFolder, 'dir')
  % If that folder doesn't exist, just start in the current folder.
  startingFolder = pwd;
end
defaultFileName = fullfile(startingFolder, '*.jpg');

[baseFileName, folder] = uigetfile(defaultFileName, 'Select a file');
if baseFileName == 0
  % User clicked the Cancel button.
  return;
end
fullFileName2 = fullfile(folder, baseFileName)
original = imread(fullFileName2);

i1 = imread(fullFileName);
i2 = imread(fullFileName2);
i3 = xor(i1, i2);
imagesc(i3);
d = sum(i3(:)) / numel(i3);

This code seems right way but red points are not at the true positions and i prefer having the result on the input images not black background.

Does anyone have an idea on the subject?

[KALYAN ACHARJYA]

Can you clarify the question again? What is the input, what result are you expecting? Don't provide the confusing data and terms.

[Murat Kocaman]

Yes inputs are the first 2 images:

First image:

Second image:

Result expecting:

I would like to mark matched blue pixels at these 2 input images with red marker.

I hope it is clear now.

Answer 1

[Image Analyst]

I'm not sure what you mean by "matched". Like you said, the colors are different. So I guess you'll have to define how close can they be and still be considered "matched". To do that you need to compute the Delta E color difference. For that, use my Delta E program here: https://www.mathworks.com/matlabcentral/fileexchange/31118-color-segmentation-by-delta-e-color-difference

You could also do a pixel-by-pixel delta E. Simply use rgb2lab() on each image, then compute the square root of the sum of the squares of the L, A, and B channels. Then threshold at some value to discriminate between "matched" and "not a close match".

[Murat Kocaman]

Thank you Image Analyst I will try this.

BTW what I mean is to check if the object has phosphorescence. Actually, I thought on working on rgb2hsv possibility because I think the images have better and exact values on that case.

I will comment here again after trying your recommendation.

[Image Analyst]

You can simply just check the mean, or max, intensity value after you turn off the lights. If there is phosphorescence, it will be higher than if there is not.

[Murat Kocaman]

I couldn't do that. I decided to go on the code I used before. Here, the result is very similar what I need.

I only need to show the result on first or second image. How to do that?

With the current code, I have the result:

How could I show these red marks on one of the 2 input images not black background?

[Image Analyst]

If you want to do it in the overlay, see this link http://www.mathworks.com/matlabcentral/answers/104889-how-to-overlay-a-foreground-image-over-a-background-image

If you want it burned into an image, you can copy the target image,

outputRgbImage = rgbImage;

then extract the color channels

% Extract the individual red, green, and blue color channels.
redChannel = rgbImage(:, :, 1);
greenChannel = rgbImage(:, :, 2);
blueChannel = rgbImage(:, :, 3);

Then apply the mask

redChannel(mask) = 255;
greenChannel(mask) = 0;
blueChannel(mask) = 0;

Then recombine:

% Recombine separate color channels into a single, true color RGB image.
outputRgbImage = cat(3, redChannel, greenChannel, blueChannel);

[Murat Kocaman]

I tried your recommendation:

startingFolder = 'C:\Users\Development\Desktop\dneme';
if ~exist(startingFolder, 'dir')
  % If that folder doesn't exist, just start in the current folder.
  startingFolder = pwd;
end
defaultFileName = fullfile(startingFolder, '*.jpg');
[baseFileName, folder] = uigetfile(defaultFileName, 'Select a file');
if baseFileName == 0
  % User clicked the Cancel button.
  return;
end
fullFileName = fullfile(folder, baseFileName)
startingFolder = 'C:\Users\Development\Desktop\dneme';
if ~exist(startingFolder, 'dir')
  % If that folder doesn't exist, just start in the current folder.
  startingFolder = pwd;
end
defaultFileName = fullfile(startingFolder, '*.jpg');
[baseFileName, folder] = uigetfile(defaultFileName, 'Select a file');
if baseFileName == 0
  % User clicked the Cancel button.
  return;
end
fullFileName2 = fullfile(folder, baseFileName)
A = imread(fullFileName);
B = imread(fullFileName2);
C = xor(A, B);
imagesc(C);
D = sum(C(:)) / numel(C);

E = A;

redChannel = A(:, :, 1);
greenChannel = A(:, :, 2);
blueChannel = A(:, :, 3);
mask = A.createMask();
redChannel(mask) = 255;
greenChannel(mask) = 0;
blueChannel(mask) = 0;
% Recombine separate color channels into a single, true color RGB image.
E = cat(3, redChannel, greenChannel, blueChannel);
imagesc(E);

I had "Dot indexing is not supported for variables of this type".

Could you see the what is wrong here?

[Image Analyst]

Why do you think matrix A is either a structure, like MATLAB is assuming from its error, or a method of a class? There is no createMask property, field, or method of A. I'm certain that part was not my suggestion and was something you modified.

[Murat Kocaman]

Yes I added a mask definition to what you had adviced. I have already tried some other mask definitions but the result was the same.

So should I change that line and find appropriate masking or what?

[Image Analyst]

I suggest you go back to my demo and look at creating a mask from thresholding the delta E image. If you insist on doing the segmentation in RGB color space, then threshold one of the color channels.