can someone help me with dithering ? I'm trying to do random dithering, I want to do halftone and pattern too, but I'm stuck at random

global selectedImage

imread(selectedIma);

X = imread(selectedImage); 
%Val = zeros(numel(X));    % initialize pdf values storing variable with zeros
k = 1; % counter variable in for loop
% loop of (3 RGB colors)
for r = 1:size(X, 1)    % for number of rows of the image
    for c =  1:size(X, 2)    % for number of columns of the image
        if(getPixel(r,c)> rand*255)
            setPixel(r,c, 255);
        else   
            setPixel(r,c, 0);
            k = k+1;    % increment counter loop 
        end
    end
end
axes(handles.axes9);
imshow(selectedImage);

Answer 1

[DGM]

I don't know what your setPixel() and getPixel() functions are, or why you're incrementing a counter variable that's apparently not used for anything. It's also unclear whether you're trying to process color images, since the comment suggests yes, but the code suggests no. If you're trying to incorporate color into a dithering process, there are trivial ways that generally produce poor results, and there are more significantly more complicated ways (that I don't have direct examples for).

You might want to start here. This discusses basic noise dithers, halftoning, ordered dithers and error diffusion dithers.

http://www.crisluengo.net/index.php/archives/355

The code that Cris uses is largely MATLAB, but some of the tools are from DIPimage.

https://www.mathworks.com/matlabcentral/fileexchange/68622-dipimage

MIMT (also on the FEX) has functions for doing various dithers, including some of those listed above. Other than some input/output handling and avoidance of DIPimage and IPT dependencies, I don't recall that the core code is much different.

% a grayscale image
inpict = imread('lena.tif'); 
inpict = rgb2gray(inpict);

% use a small sample to reduce display interpolation effects
inpict = inpict(128:384,128:384);

% noise dithers
Dwhite = noisedither(inpict,'white');
Dblue = noisedither(inpict,'blue');

% basic ordered dither
Dordered = orddither(inpict);

% Zhou-Fang variable-coefficient error-diffusion dither
Derrdiff = zfdither(inpict);

Regarding color, the simple way is to just do the comparison on each of the three RGB channels independently. This results in an 8-color image that's generally pretty terrible. The only colors you get are black, white, and 6 primary and secondary colors. You can't apply a colormap or anything. This also kind of defeats the point of error diffusion. For an example of how the simple method is done, you can look at how orddither() does it (it's just a simple loop at the end).

Dordered = orddither(inpict,16,'color');

This isn't the method used by rgb2ind() with the 'dither' flag. That requires color quantization and it requires the index assignment to be incorporated into the error diffusion process. I don't have an example for that.

Due to a difference in intent and scope, the MIMT halftoning tools are more complicated (and cumbersome) than what Cris describes in the blog. You'd be better off just using his examples.

[suzan salem]

thank you for answering but i want a code for the function itself and how it dithers, not using a predfined function for dithering, it's for an asignment with a specfied descrptiong on how to do it but i am still lost, could you help me with that?

[DGM]

The blog and the functions themselves serve as example code. The fact that the functions are in a working form already means that you can look at how they work. They don't really have much internal comments, but if there's some particular point of confusion, I can try to elaborate. If you have a start on your own versions of the code and are having problems, I can try to straighten that out. You'll just need to let me know what specifically you need help with.

EDIT: I went ahead and added some comments to the function files that might be helpful. I don't know if you need to bother with the error-diffusion dithering, but for the noise dither, it should be fairly clear. You probably don't need to do certain things for your assignment, like generating the Bayer matrix for the ordered dithering method. You can just use a literal precalculated matrix like Cris does. If you don't need to deal with color, you can avoid a few lines of code as well.

Some of the functions used by these files are parts of MIMT. You should be able to easily avoid them.

MIMT imfilterFB() can be replaced directly with imfilter() from IPT

MIMT adapthisteqFB() can be replaced directly with adapthisteq() from IPT

MIMT imcast() can be avoided by using the appropriate conversion tool such as im2double(), im2uint8(), etc

MIMT simnorm() can be replaced with mat2gray() in these cases

Cases where MIMT mono() is used to calculate luma:

img = mono(img,'y');

can be replaced with

img = rgb2gray(img);

... I think that should be most of it.

[suzan salem]

thanks! one more question, my asignment has a "pattern dithering" is it like bayer ordered dithering or floyed? and given us a 3x3 matrix for the pattern dithering.

[DGM]

I'm going to assume that your "pattern dithering" is synonymous with a Bayer/ordered dither. Floyd-Steinberg is a basic error-diffusion dither.

I suppose you could use a 3x3 matrix in an ordered dither. You can really use any pattern matrix you want. The standard Bayer matrix is just a means of getting certain uniform dither patterns to arise from uniform gray regions in an image. For what it's worth, this is what would happen if your pattern matrix were [0 1 2 3 4 5 6 7].'/8

Once you have your matrix, just replicate it to cover the image area, trim to size, and then perform the thresholding.

[suzan salem]

can i do it using this matrix?

8 3 4

6 1 2

7 5 9

[DGM]

This is just pulled straight out of orddither.m and turned into a script:

% a grayscale image
inpict = imread('lena.tif'); 
inpict = rgb2gray(inpict);
% use a small sample to reduce display interpolation effects
inpict = inpict(128:384,128:384);

% convert to unit-scale double
inpict = im2double(inpict);
sz = size(inpict);
numchans = size(inpict,3);

% generate the coefficient matrix
B = [8 3 4; 6 1 2; 7 5 9];
B = (B-1)/9; % normalize the coefficient matrix

% create mask by tiling the coefficient matrix 
% so that it covers the same area as the image
n = size(B);
t = repmat(B,ceil(sz./n));
t = t(1:sz(1),1:sz(2));

% binarize image by comparing gray levels to tiled mask
out = inpict > t;

imshow(out)