How can I convert an RGB image to grayscale but keep one color?

I am trying to create an effect similar to Sin City or other movies where they remove all colors except one from an image.

I have an RGB image which I want to convert to grayscale but I want to keep one color.

This is my picture:

I want to keep the red color. The rest should be grayscale.

This is what my code outputs so far (you can see that the areas are correct, I don't know why they are white instead of red though):

Here is my code so far:

filename = 'roses.jpg';

[cdata,map] = imread( filename );
% convert to RGB if it is indexed image
if ~isempty( map ) 
   cdata = idx2rgb( cdata, map ); 
end

%imtool('roses.jpg');

imWidth = 685;
imHeight = 428;

% RGB ranges of a color we want to keep
redRange = [140 255];
greenRange = [0 40];
blueRange = [0 40];

% RGB values we don't want to convert to grayscale
redToKeep = zeros(imHeight, imWidth);
greenToKeep = zeros(imHeight, imWidth);
blueToKeep = zeros(imHeight, imWidth);

for x=1:imWidth

    for y=1:imHeight

        red = cdata( y, x, 1 );
        green = cdata( y, x, 2 );
        blue = cdata( y, x, 3 );

        if (red >= redRange(1) && red <= redRange(2) && green >= greenRange(1) && green <= greenRange(2) && blue >= blueRange(1) && blue <= blueRange(2))
            redToKeep( y, x ) = red;
            greenToKeep( y, x ) = green;开发者_如何学运维
            blueToKeep( y, x ) = blue;
        else
            redToKeep( y, x ) = 999;
            greenToKeep( y, x ) = 999;
            blueToKeep( y, x ) = 999;
        end

    end 

end 

im = rgb2gray(cdata);
[X, map] = gray2ind(im);
im = ind2rgb(X, map);

for x=1:imWidth

    for y=1:imHeight

        if (redToKeep( y, x ) < 999)
            im( y, x, 1 ) = 240;
        end
        if (greenToKeep( y, x ) < 999)
            im( y, x, 2 ) = greenToKeep( y, x );
        end
        if (blueToKeep( y, x ) < 999)
            im( y, x, 3 ) = blueToKeep( y, x );
        end

    end 

end 

imshow(im);

---

One option which greatly improves the quality of the resulting image is to convert to a different color space in order to more easily select your colors. In particular, the HSV color space defines pixel colors in terms of their hue (the color), saturation (the amount of color), and value (the brightness of the color).

For example, you can convert your RGB image to HSV space using the function rgb2hsv, find pixels with hues that span what you want to define as "non-red" colors (like, say, 20 degrees to 340 degrees), set the saturation for those pixels to 0 (so they are grayscale), then convert the image back to RGB space using the function hsv2rgb:

cdata = imread('EcyOd.jpg');       % Load image
hsvImage = rgb2hsv(cdata);         % Convert the image to HSV space
hPlane = 360.*hsvImage(:, :, 1);   % Get the hue plane scaled from 0 to 360
sPlane = hsvImage(:, :, 2);        % Get the saturation plane
nonRedIndex = (hPlane > 20) & ...  % Select "non-red" pixels
              (hPlane < 340);
sPlane(nonRedIndex) = 0;           % Set the selected pixel saturations to 0
hsvImage(:, :, 2) = sPlane;        % Update the saturation plane
rgbImage = hsv2rgb(hsvImage);      % Convert the image back to RGB space

And here is the resulting image:

Notice how, compared to the solution from zellus, you can easily maintain the light pink tones on the flowers. Notice also that brownish tones on the stem and ground are gone as well.

For a cool example of selecting objects from an image based on their color properties, you can check out Steve Eddins blog post The Two Amigos which describes a solution from Brett Shoelson at the MathWorks for extracting one "amigo" from an image.

A note on selecting color ranges...

One additional thing you can do which can help you select ranges of colors is to look at a histogram of the hues (i.e. hPlane from above) present in the pixels of your HSV image. Here's an example that uses the functions histc (or the recommended histcounts, if available) and bar:

binEdges = 0:360;    % Edges of histogram bins
hFigure = figure();  % New figure

% Bin pixel hues and plot histogram:
if verLessThan('matlab', '8.4')
  N = histc(hPlane(:), binEdges);  % Use histc in older versions
  hBar = bar(binEdges(1:end-1), N(1:end-1), 'histc');
else
  N = histcounts(hPlane(:), binEdges);
  hBar = bar(binEdges(1:end-1), N, 'histc');
end

set(hBar, 'CData', 1:360, ...            % Change the color of the bars using
          'CDataMapping', 'direct', ...  %   indexed color mapping (360 colors)
          'EdgeColor', 'none');          %   and remove edge coloring
colormap(hsv(360));                      % Change to an HSV color map with 360 points
axis([0 360 0 max(N)]);                  % Change the axes limits
set(gca, 'Color', 'k');                  % Change the axes background color
set(hFigure, 'Pos', [50 400 560 200]);   % Change the figure size
xlabel('HSV hue (in degrees)');          % Add an x label
ylabel('Bin counts');                    % Add a y label

And here's the resulting pixel color histogram:

Notice how the original image contains mostly red, green, and yellow colored pixels (with a few orange ones). There are almost no cyan, blue, indigo, or magenta colored pixels. Notice also that the ranges I selected above (20 to 340 degrees) do a good job of excluding most everything that isn't a part of the two large red clusters at either end.

---

figure
pic = imread('EcyOd.jpg');

for mm = 1:size(pic,1)
    for nn = 1:size(pic,2)
        if pic(mm,nn,1) < 80 || pic(mm,nn,2) > 80 || pic(mm,nn,3) > 100
            gsc = 0.3*pic(mm,nn,1) + 0.59*pic(mm,nn,2) + 0.11*pic(mm,nn,3);
            pic(mm,nn,:) = [gsc gsc gsc];
        end
    end
end
imshow(pic)

---

I don't really know how matlab works so I can't really comment on the code, but maybe this will help explain a bit how RGB colors work.

When using RGB colors a gray scale can be made by making sure the values for R,G and B are all the same. So basically what you want to do is detect if a pixel is red, when not just make R,G and B the same (you can use an average of the 3 for a rudimentary result).

Harder part is how to detect if a pixel is actually red, you can't just check if a pixel is high in the R value since it can still be another color, and a low R value can just mean a darker red.

so you could do something like this: (I don't have matlab, so assuming syntax):

red = cdata( y, x, 1 );
green = cdata( y, x, 2 );
blue = cdata(y, x, 3);

if (red < (blue * 1.4) || red < (green * 1.4) )
{
    avg = (red + green + blue) / 3;
    cdata(y, x, 1) = avg;
    cdata(y, x, 2) = avg;
    cdata(y, x, 3) = avg;
}

There are probably better ways to detect red and to get an average gray, but it's a start ;)