# How can i use threshold to convert a gray-scaled image into binary image ?

75 views (last 30 days)
Christine Ak on 19 Nov 2013
Commented: DGM on 25 Aug 2023
How can I use threshold to convert a gray-scaled image into binary image , I mean to get the image just black and white ?
Does Binarization help ??
Thx

Jan on 19 Nov 2013
Edited: Jan on 19 Nov 2013
This is straightforward:
A = imread('cameraman.tif'); % example grayscale image
threshold = 120; % custom threshold value
A_bw = A > threshold;
=======
edit: removed factor of 255 as Image Analyst pointed out
Christine Ak on 19 Nov 2013
Thxxxx :d

Simon on 19 Nov 2013
Hi!
If you load an image into matlab, you get a matrix A (for example) of size (XxYx3) with X and Y being the number of pixels in x- and y-direction. Usually this matrix is for RGB images. Look at imread
If it is a grayscale image the values for all three colors are the same, they range between 0 and 255. You may now apply a threshold
% threshold
t = 128;
% find values below
ind_below = (A < t);
% find values above
ind_above = (A >= t);
% set values below to black
A(ind_below) = 0;
% set values above to white
A(ind_above) = 255;
Image Analyst on 23 Mar 2023
@Nayana again (see my answer) there is no need to do all that and make A a double matrix of 0 and 255. You can simply do
And of course he messes up the size of A. It's not "of size (XxYx3)". It's of size YxXx3 which is rows x columns x 3.
DGM on 25 Aug 2023
(i don't remember why i parked my browser on this page, but I'll bite anyway)
% find values below
ind_below = (A < t);
% find values above
ind_above = (A >= t);
This whole baloney is unnecessary. Do the comparison once. If you need, negate the logical result, that way you know the union of masks spans the entire input. Comparing it twice just wastes time and invites room for error. That said, there's no need to do so at all.
A logical image as would result from a comparison operator (e.g. >=) is a properly-scaled image itself. It can be rescaled for viewing or saving without problem. Creating a floating-point image in uint8-scale, or otherwise blindly presuming that the input is uint8 is to create an improperly-scaled image which won't display or save correctly without intervention. Learn how images are scaled according to their numeric class.
There's no reason to create two masks. There's no reason to overwrite the input. There's no reason to need two operations to do so. Stop creating problems for yourself. As IA and Jan have posted, this whole thing reduces to a single line of code, and that single line of code is more robust than this answer.

Image Analyst on 19 Nov 2013
Yes you binarize the image by thresholding:
binaryImage = grayImage > thresholdValue;
There is no need to ever multiply by 255 that I've ever encountered. Displaying the binary (logical) image will show it as black and white even without multiplying by, or directly setting to, a value of 255.
Jan on 19 Nov 2013
Right, in general (and especially for binarized images) the scaling is obsolete...
Christine Ak on 19 Nov 2013
Thxxx alot

burçin temur on 28 May 2020
how can I change the red band with the green band in image?
##### 3 CommentsShow 1 older commentHide 1 older comment
burçin temur on 28 May 2020
thanks
DGM on 12 Feb 2023
Or just:
rgbImage = rgbImage(:,:,[2 1 3]);

Ali nafaa on 29 Nov 2022
Edited: Image Analyst on 29 Nov 2022
figure,imshow(x);
[r,c] = size (x);
output=zeros(r,c);
for i = 1 : r
for j = 1 : c
if x(i,j) > 128
output(i,j)=1;
else
output(i,j)=0;
end
end
end
figure,imshow(output);
Image Analyst on 29 Nov 2022
Sorry, I changed it to x and output like you used. I find it's easier for people to understand the code if you use descriptively named variables. Usually people think of x as like in an x-y graph, not a binary image. So I'd rather use "mask" or "binaryImage" rather than "output", and "grayImage" rather than "x".
Ali nafaa on 29 Nov 2022
yes thank you