Colour pattern reading using only Lab or HSV.

I have an image with squres with different colours: red, green, blue, yellow and white (see attached image). Im wondering if it possible to select a group of pixels on each individual square and for the selected pixels to represent the colour and producing an array of 4x4 matrix.

[DGM]

See also:

https://www.mathworks.com/matlabcentral/answers/1664934-returning-an-array-of-colors-from-a-double-image#answer_912294

https://www.mathworks.com/matlabcentral/answers/454146-how-to-detect-each-color-square-of-an-image-to-crop-as-individual-images

Answer 1

[Image Analyst]

I would convert to HSV color space and then threshold the S channel to get the squares. Then use the mask and regionprops() to get the meanIntensity of each square. Here's a start/hint:

hsvImage = rgb2hsv(rgbImage);
mask = hsvImage(:, :, 2) > 0.25; % Only consider really vivid pixels, not gray ones.
mask = bwareaopen(mask, 1000); % Remove noise.
% Extract the individual red, green, and blue color channels.
redChannel = rgbImage(:, :, 1);
greenChannel = rgbImage(:, :, 2);
blueChannel = rgbImage(:, :, 3);
% Get red means
props = regionprops(mask, redChannel, 'MeanIntensity');
allMeansR = [props.MeanIntensity];
% Get green means
props = regionprops(mask, greenChannel, 'MeanIntensity');
allMeansG = [props.MeanIntensity];
% Get red means
props = regionprops(mask, blueChannel, 'MeanIntensity');
allMeansB = [props.MeanIntensity];

[Stephen Otieno]

Hi image Analyst, Im not quiet confident using hsv and would like a clarification on "then threshold the s channel to get squares". Currenly I have been using Lab and I have attached my code(2), I've so far been able to select each colour and Im wonder if you've got suggestion on what my next step will be.

Answer 2

[Image Analyst]

OK Stephen, sorry for the delay - you've probably already solved it by now, but for what it's worth, here's a full blown demo for you:

% Demo to mask square tiles and measure their mean RGB from an RGB image.  By Image Analyst, April 16, 2020.
clc;    % Clear the command window.
close all;  % Close all figures (except those of imtool.)
clearvars;
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;
fontSize = 16;

% Read in image.
folder = pwd;
baseFileName = 'org_1.png';
fullFileName = fullfile(folder, baseFileName);
rgbImage = imread(fullFileName);
[rows, columns, numberOfColorChannels] = size(rgbImage);
% Display the original image.
h1 = subplot(1, 2, 1);
imshow(rgbImage, []);
axis('on', 'image');
caption = sprintf('Original Color Image\n"%s"', baseFileName);
title(caption, 'FontSize', fontSize, 'Interpreter', 'None');
drawnow;
hp = impixelinfo(); % Set up status line to see values when you mouse over the image.

% Set up figure properties:
% Enlarge figure to full screen.
hFig = gcf;
hFig.Units = 'Normalized';
hFig.WindowState = 'maximized';
% Get rid of tool bar and pulldown menus that are along top of figure.
% set(gcf, 'Toolbar', 'none', 'Menu', 'none');
% Give a name to the title bar.
hFig.Name = 'Demo by Image Analyst';
hFig.NumberTitle = 'Off'

hsvImage = rgb2hsv(rgbImage);
mask = hsvImage(:, :, 2) > 0.25; % Only consider really vivid pixels, not gray ones.
% Fill noise holes
mask = imfill(mask, 'holes');
% Get the areas
% props = regionprops(mask, 'Area');
% areas = sort([props.Area], 'Descend')
% Take the 16 biggest blobs:
mask = bwareafilt(mask, 16);
mask = bwareaopen(mask, 1000); % Remove noise.
% Display the image.
h2 = subplot(1, 2, 2);
imshow(mask);
axis('on', 'image');
title('Mask Image of Square Tiles', 'FontSize', fontSize, 'Interpreter', 'None');
drawnow;
hp = impixelinfo(); % Set up status line to see values when you mouse over the image.

% Extract the individual red, green, and blue color channels.
redChannel = rgbImage(:, :, 1);
greenChannel = rgbImage(:, :, 2);
blueChannel = rgbImage(:, :, 3);
% Get red means
propsR = regionprops(mask, redChannel, 'MeanIntensity', 'Centroid');
allMeansR = [propsR.MeanIntensity];
% Get green means
propsG = regionprops(mask, greenChannel, 'MeanIntensity', 'Centroid');
allMeansG = [propsG.MeanIntensity];
% Get red means
propsB = regionprops(mask, blueChannel, 'MeanIntensity', 'Centroid');
allMeansB = [propsB.MeanIntensity];

% Print the values
for k = 1 : length(propsR)
	caption = sprintf('Tile %d', k);
	% First on the color image.
	text(h1, propsR(k).Centroid(1), propsR(k).Centroid(2), caption, ...
		'Color', 'k', 'FontSize', 20, 'HorizontalAlignment', 'center');
	% Then on the mask
	text(h2, propsR(k).Centroid(1), propsR(k).Centroid(2), caption, ...
		'Color', 'b', 'FontSize', 20, 'HorizontalAlignment', 'center');
	
	fprintf('For Tile #%2.2d, Mean RGB = (%.3f, %.3f, %.3f).\n', ...
		k, allMeansR(k), allMeansG(k), allMeansB(k));
end
msgbox('Done!');

For Tile #01, Mean RGB = (244.618, 57.526, 21.911).

For Tile #02, Mean RGB = (244.867, 57.693, 21.614).

For Tile #03, Mean RGB = (244.857, 229.460, 43.958).

For Tile #04, Mean RGB = (43.470, 15.524, 245.138).

For Tile #05, Mean RGB = (244.845, 57.584, 21.808).

For Tile #06, Mean RGB = (152.526, 244.603, 124.558).

For Tile #07, Mean RGB = (152.184, 244.773, 124.323).

For Tile #08, Mean RGB = (152.723, 244.814, 124.676).

For Tile #09, Mean RGB = (43.232, 15.108, 244.669).

For Tile #10, Mean RGB = (244.858, 229.398, 44.333).

For Tile #11, Mean RGB = (43.118, 15.633, 245.012).

For Tile #12, Mean RGB = (245.048, 57.044, 21.543).

For Tile #13, Mean RGB = (43.727, 15.511, 245.031).

For Tile #14, Mean RGB = (244.918, 57.757, 21.881).

For Tile #15, Mean RGB = (42.992, 15.314, 244.726).

For Tile #16, Mean RGB = (42.924, 15.437, 244.618).

Note the labeling order is in the usual way blobs are ordered: top down, then from left to right. So the ordering might not be what you expect because some squares have a little spur (due to noise) that sticks out on the left side which might cause it to be encountered and measured before a square above it.

[Stephen Otieno]

Thanks for your help image analyst.

I have similar images that are distored and im wondering if you've any tips or demos on how to correct the image back to its original shape.

[Image Analyst]

I'd first segment the black spots, then use imwarp() to put the spots in known locations in the corners. I don't have demo code for that so you're on your own. You might try this link.