MATLAB: Drawing a line over a black and white image

Question

What is the best way to draw a line over a black and white (binary) image in MATLAB, provided the start and end coordinates are known?

Please note, I am not trying to add an annotation line. I would like the line to become part of the image.

Solution 1

You may want to look at my answer to an SO question about adding a line to an image matrix. Here's a similar example to the one I have in that answer, which will make a white line running from row and column index (10, 10) to (240, 120):

img = imread('cameraman.tif');  % Load a sample black and white image
x = [10 240];                   % x coordinates
y = [10 120];                   % y coordinates
nPoints = max(abs(diff(x)), abs(diff(y)))+1;    % Number of points in line
rIndex = round(linspace(y(1), y(2), nPoints));  % Row indices
cIndex = round(linspace(x(1), x(2), nPoints));  % Column indices
index = sub2ind(size(img), rIndex, cIndex);     % Linear indices
img(index) = 255;  % Set the line points to white
imshow(img);       % Display the image

And here's the resulting image:

Solution 2

If you are bothered by exceptional cases of other methods here's a bullet-proof method that results in a line:

• whose pixels always touch each other during the whole length of the line (pixels are 8-neighbors to each other),
• density of the line is not dependent on the additional parameter, but is determined flexibly to accommodate guarantee from the first point.

---

Inputs (convenient for making function out of this code):

• img - matrix that contains image,
• x1, y1, x2, y2 - coordinates of the end points of the line to be drawn.

Code:

% distances according to both axes
xn = abs(x2-x1);
yn = abs(y2-y1);

% interpolate against axis with greater distance between points;
% this guarantees statement in the under the first point!
if (xn > yn)
    xc = x1 : sign(x2-x1) : x2;
    yc = round( interp1([x1 x2], [y1 y2], xc, 'linear') );
else
    yc = y1 : sign(y2-y1) : y2;
    xc = round( interp1([y1 y2], [x1 x2], yc, 'linear') );
end

% 2-D indexes of line are saved in (xc, yc), and
% 1-D indexes are calculated here:
ind = sub2ind( size(img), yc, xc );

% draw line on the image (change value of '255' to one that you need)
img(ind) = 255;

---

Here's the example image with three lines drawn on it:

Solution 3

This algorithm offers one approach.

Solution 4

It actually is just a modification on plesiv's answer. I'm drawing thousands of lines over an image and I need to increase the performance. The most improvement made by omitting interp1 calls and using integer variables made it slightly faster. It performs about 18% faster on my PC comparing to plesiv's code.

function img = drawLine(img, x1, y1, x2, y2)
x1=int16(x1); x2=int16(x2); y1=int16(y1); y2=int16(y2);
% distances according to both axes
xn = double(x2-x1);
yn = double(y2-y1);

% interpolate against axis with greater distance between points;
% this guarantees statement in the under the first point!
if (abs(xn) > abs(yn))
    xc = x1 : sign(xn) : x2;
    if yn==0
        yc = y1+zeros(1, abs(xn)+1, 'int16');
    else
    yc = int16(double(y1):abs(yn/xn)*sign(yn):double(y2));
    end
else
    yc = y1 : sign(yn) : y2;
    if xn==0
        xc = x1+zeros(1, abs(yn)+1, 'int16');
    else
    xc = int16(double(x1):abs(xn/yn)*sign(xn):double(x2));
    end
end

% 2-D indexes of line are saved in (xc, yc), and
% 1-D indexes are calculated here:
ind = sub2ind(size(img), yc, xc);

% draw line on the image (change value of '255' to one that you need)
img(ind) = 255;
end

View more solutions

Comments

[Jonas]

This works perfectly for a diagonal line, but may add unwanted pixels for a flatter line. If you don't care about the additional pixels, I suggest choosing gnovices solution because it is fast and simple.

[taranaki]

It would be nice for this method to gracefully handle out-of-bounds points. I used some ray-intersection code to find the bounds given a point-slope form of a line. It's done before calling this function but it could easily be incorporated instead.

[plesiv]

Yeah... Sorry. I wrote this long time ago. It takes too much effort to make improvements now, since I don't have Matlab currently.

[gnovice]

@Jonas: I've updated the algorithm to compute the line better, which removes some unnecessary pixels.

[Jonas]

Thank you for improving my answer!

[saastn]

insertShape is really slow and it only returns images in RGB format.

[saastn]

It can be even faster if you remove interp1 calls, considering your comment above I posted a new answer.