How to extract a particular colour from an RGB image?

Dear all,

I want to extract a particular colour from an RGB image. If that specific colour is present in an image, I wish to replace it with a zero pixel intensity. In the figure below, the arrow points out the light red colour; here, I want to extract this particular light red colour from the image. If it is present in the image, I want to replace it with a zero intensity level with that particular colour.

How to make a logic to extract a particular colour from an image?

Answer 1

[Walter Roberson]

Not perfect, but it will give you a starting point.

filename = 'https://www.mathworks.com/matlabcentral/answers/uploaded_files/641315/f00167.png';
rgb = imread(filename);
HSV = rgb2hsv(rgb);
mask1 = HSV(:,:,1) < 0.2;
mask2 = HSV(:,:,2) > 0.3 & HSV(:,:,2) < 0.7;
mask3 = HSV(:,:,3) > 0.2 & HSV(:,:,3) < 0.6;
mask = mask1 & mask2 & mask3;
BW = bwareafilt(mask,[30 inf]);
filtered_RGB = rgb .* uint8(BW);
imshow(filtered_RGB)

[Image Analyst]

And to replace the red with black, you can do this to mask the image to black wherever the BW binary image is true (indicating red color in the scene):

% Mask the image using bsxfun() function to multiply the mask by each channel individually.  Works for gray scale as well as RGB Color images.
maskedRgbImage = bsxfun(@times, rgbImage, cast(~BW, 'like', rgbImage));

[Walter Roberson]

To blacken out the red, you can use a minor change to what I wrote before

filename = 'https://www.mathworks.com/matlabcentral/answers/uploaded_files/641315/f00167.png';
rgb = imread(filename);
HSV = rgb2hsv(rgb);
mask1 = HSV(:,:,1) < 0.2;
mask2 = HSV(:,:,2) > 0.3 & HSV(:,:,2) < 0.7;
mask3 = HSV(:,:,3) > 0.2 & HSV(:,:,3) < 0.6;
mask = mask1 & mask2 & mask3;
BW = bwareafilt(mask,[30 inf]);
filtered_RGB = rgb .* uint8(~BW);
imshow(filtered_RGB)

Looks like the algorithm needs some tuning, to get the top of the curb as well.

[SUKUMAR NAGINENI]

Thank you, @Walter Roberson and @Image Analyst, for your insightful suggestions.

Dear @Walter Roberson

This code will be helpful. But, some light red colour is present in the above image; it was not replaced with the black pixels. I am interested to know that how you have selected these values like 0.2, 0.3, 0.7... I want to repeat this code for other images, which contains a light red colour. In case if I selected some different colour, for that, how to choose these values. Please give me suggestions.

mask1 = HSV(:,:,1) < 0.2;

mask2 = HSV(:,:,2) > 0.3 & HSV(:,:,2) < 0.7;

mask3 = HSV(:,:,3) > 0.2 & HSV(:,:,3) < 0.6;

mask = mask1 & mask2 & mask3;

BW = bwareafilt(mask,[30 inf]);

[Walter Roberson]

The way I selected the values was that I used

filename = 'https://www.mathworks.com/matlabcentral/answers/uploaded_files/641315/f00167.png';
rgb = imread(filename);
HSV = rgb2hsv(rgb);
figure(1); imshow(HSV(:,:,1)); title('H');
figure(2); imshow(HSV(:,:,2)); title('S');
figure(3); imshow(HSV(:,:,3)); title('V');

and then I used datatips to look at the values in the area of interest and compare them to other values in HSV representation.

It is not possible to just ask to filter "light red" because "light red" does not have a specific meaning. Humans automatically change perceived color based upon surrounding contrast and edges and perception of what kind of light is present in an area. For example if you are standing under a sodium arc-lamp streetlight, then to an extent you will continue to perceive things you know to be red, as being red, even though if you took a photo and checked the RGB, the pixels would be far from something you would normally call "red".

[SUKUMAR NAGINENI]

I understood it clearly

Thank you @Walter Roberson and @Image Analyst

[Walter Roberson]

(My internal tracking number for this matter is T0098902 )

[Image Analyst]

You can use the Color Thresholder app on the Apps tab of the tool ribbon if you'd like to interactively adjust the thresholds.

You can "Thank" Walter by accepting the answer and/or voting for it.

[SUKUMAR NAGINENI]

Sorry, I forgot to click on the accepted answer.

Thank you @Walter Roberson and @Image Analyst

[Rupika]

Hi, I am doing something similar to this. After obtainig the filtered_RGB image, i want to blend only the colored parts to another image on a particular location. But when i tried it the backgound black region is also getting blended. Can you help me!!

Answer 2

[Sulaymon Eshkabilov]

(1) Image read: imread('traffic.png')

(2) Select and Find out the pixel value ranges of a red color, e.g.:

I = imread('traffic.png');
PXY = impixel(I);   % Once done selecting enough number of pixel points, hit ENTER from keyboard
R = (PXY(:,1));     % G = (PXY(:,2)); B = (PXY(:,3)); % If interested in Green Or Blue 
Rm=min(R);          % Gm=min(G); Bm=min(B);

(3) Use logical indexing, e.g.:

RED = I(:,:,1);
GREEN = I(:,:,2);
BLUE = I(:,:,3);
RED(RED>=Rm)=0;     % For RED color layer
GREEN(RED>=Rm)=0;   % For GREEN color layer
BLUE(RED>=Rm)=0;    % For BLUE color layer
NEW_Im(:,:,1) =RED;
NEW_Im(:,:,2) =GREEN;
NEW_Im(:,:,3) =BLUE;
imshow(NEW_Im), shg