How to Convert data into Image using Matlab

Hello Everyone, I Hope you are doing well. I have the following data, I have written a code to convert the data into Image. But when i round the values the Image does not same as the plot, I am doing round because each value represent a pixel value.

The Dataplot.jpg shows the original data, which is also attached in dataScan.mat. The output image after the code is imagefromdataset.jpg.

Is there is any way to get the same shape as original data. Due to rounding the value the shape changes.

How can i modified the code to get the same shape in image.

%% create grayscale shapes that resemble the data
[numImages, lenImage] = size(dataset);
imSz = 1000; % assuming images are 1000x1000
imbg = false(imSz);                                  % background "color"
imfg = ~imbg(1,1);                                  % forground "color"
imSizeOut=[1000 1000];                                % ImageSize
for imNum = 1:numImages        
    imData = round(dataset(imNum,:));  % get pattern
    
    [~,Y] = meshgrid(1:imSz); % make a grid
    
    % black and white image
    BW = imbg;
    BW(Y==imData)=imfg;
   
    % resize (from 1000x1000)
    BW=imbinarize(imresize(uint8(BW),imSizeOut));
    % convert to uint8 (0 255)
    im = im2uint8(BW);
    SE=strel('disk',2);
    BW=imdilate(im,SE);
    %im = BW;
    im = flipud(BW);
end

[Rik]

Why are you doing it like this, instead of capturing the image from the plot?

I don't see what is going wrong. You picked a low resolution and a large range. What exactly is different between this result and your expectations?

[Stephen john]

@Rik I want to create a binary Image thats why i am doing this , is there is any other method Which is use to convert data into image? The wrong is when i round decimal values the shape is change

Answer 1

[Rik]

The shape did not actually change. You just expanded the y from [-2 18] to [0 1000].

S=load(websave('tmp.mat','https://www.mathworks.com/matlabcentral/answers/uploaded_files/1119810/dataScan.mat'));
dataset=S.dataset;
plot(dataset,'o'),ylim([0 1000])

If you want to reproduce this data yourself, one way to that is to create the coordinate grid explicitly:

x=1:numel(dataset);%this is what you did implictly
y=dataset;

resolution=1000;

X=linspace(min(x), max(x) ,resolution);
Y=linspace(min(y), max(y) ,resolution);

% determine index in X of each x, likewise for y (note that the coordinates
% are flipped for the y-direction in images).
ind_x=interp1(X,  1:numel(X) ,x,'nearest');
ind_y=interp1(Y,numel(Y):-1:1,y,'nearest');

%compute linear index
ind=sub2ind([resolution resolution],ind_y,ind_x);

% create the image
IM=zeros(resolution,resolution);
IM(ind)=1;

% Since you apparently want a dilation, apply it before displaying the
% image.
SE=strel('disk',2);
imshow(imdilate(IM,SE))

[Stephen john]

@Rik

When the values are fixed there is no change it gives the following error

load Dataset2
dataset=ans;
x=1:numel(dataset);%this is what you did implictly
y=dataset;
resolution=1000;
X=linspace(min(x), max(x) ,resolution);
Y=linspace(min(y), max(y) ,resolution);
% determine index in X of each x, likewise for y (note that the coordinates
% are flipped for the y-direction in images).
ind_x=interp1(X,  1:numel(X) ,x,'nearest');
ind_y=interp1(Y,numel(Y):-1:1,y,'nearest');

Error using matlab.internal.math.interp1
Sample points must be unique.

Error in interp1 (line 188)
VqLite = matlab.internal.math.interp1(X,V,method,method,Xqcol);

%compute linear index
ind=sub2ind([resolution resolution],ind_y,ind_x);
% create the image
IM=zeros(resolution,resolution);
IM(ind)=1;
% Since you apparently want a dilation, apply it before displaying the
% image.
SE=strel('disk',2);
imshow(imdilate(IM,SE))

[Rik]

Next time, please attach the mat file to the same comment that has the code and use the green triangle to run the code. I will now edit this thread.

You also don't need to @-mention me every comment.

What is the range you want the y-values to have on your image?

My code assumes you want the full data range, but if you only have a single y value, then that doesn't make much sense, as you would simply get a white image.

What do you want to happen in that case?

[Stephen john]

@Rik I have attached the dataset, Have you check on this for the error?

[Rik]

I asked you several question (which you ignored) and told you that you don't need to @-mention me every comment (which you ignored as well).

Why exactly do you expect me to respond?

[Stephen john]

The above code you share does not work on the data which i have attached below, When i run the code the error comes

[Rik]

Of course the code doesn't work. You only have 1 y-value. Do you want a completely white image with a height of 1 pixel?

You also misread the part that you now did: you attached the mat file to a comment that does not contain your code. Whenever you post code or mention an error, you should make sure we can see that error. That means you have to put the code and the data in a single comment and use the green triangle to run your code.

In this case that is not needed, because the cause is not your data or my code. The cause is that there is an edge case that you didn't explain. It is not clear what should happen. That is something you need to explain.

[Stephen john]

yeah it has only one values means we have a straight white line in the image.

The same code i used which you have pasted but it gives the above error as i show.

I need a code which works on this data as well.

[Stephen john]

the following error

Error using matlab.internal.math.interp1
Sample points must be unique.

Error in interp1 (line 188)
        VqLite = matlab.internal.math.interp1(X,V,method,method,Xqcol);

Answer 2

[Walter Roberson]

yeah it has only one values means we have a straight white line in the image.

Y=linspace(min(y), max(y) ,resolution);

With your y values all being the same, min(y) and max(y) are going to be equal, so you would be generating a constant vector with resolution elements in it.

ind_y=interp1(Y,numel(Y):-1:1,y,'nearest');

Your Y values (the independent variable for interpolation purposes) are all the same, but interp1() requries that the values of the independent variables are all different.

I need a code which works on this data as well.

Well you cannot use interp1() in that case.

Your situation is asking interp1([5 5 5], [3 2 1], [5 5 5 5], 'nearest') . Yes, 5 (in the query values) appear in the independent vector [5 5 5], but do you say that they should be considered to be come from location 3, or location 2, or location 1 ?

You will need to detect this case and treat it differently.

[Stephen john]

@Walter Roberson Anyother solution to solve this problem?

[Walter Roberson]

if Y(1) == Y(end)
    do something else
else
    proceed with the interp1()
end

but in the "do something else" section you need to figure out what a reasonable output would be.

I am not clear at the moment as to why you are not using Computer Vision insertShape() to "draw" the line into a matrix without any of that interp1() stuff.

[Stephen john]

@Walter Roberson Is there is anyother code to draw this lines and shape into binary image?

[Walter Roberson]

Computer Vision insertShape() makes it easy.

But if you do not have the Computer Vision Toolbox, then use Bresenham's Line Algorithm https://en.wikipedia.org/wiki/Bresenham%27s_line_algorithm

see

https://www.mathworks.com/matlabcentral/fileexchange/28190-bresenham-optimized-for-matlab or https://www.mathworks.com/matlabcentral/fileexchange/25544-line-drawing-by-bresenham-algorithm

[Stephen john]

@Walter Roberson Can you provide any solution for this?

[Walter Roberson]

Scale your coordinates to the size of the output image, x to fit OUTPUT_COLUMNS and y to fit OUTPUT_ROWS

Coords = [Scaled_x(:), Scaled_y(:)];
M = zeros(OUTPUT_ROWS, OUTPUT_COLUMNS);
M = insertShape(M, 'line', Coords, 'Color', 'b');  %outputs RGB

M = M(:,:,1) ~= 0;  %convert RGB to binary

[Stephen john]

@Walter Roberson It will not work on whole dataset

[Walter Roberson]

What error message are you getting?

OUTPUT_ROWS = 384; OUTPUT_COLUMNS = 384;
Scaled_x = [1:200, 200, 200:-1:150, 150, 150:175];
Scaled_y = [50*ones(1,200), 150, 151:151+51-1, 83, 84:-1:59];

Coords = [Scaled_x(:), Scaled_y(:)];
M = zeros(OUTPUT_ROWS, OUTPUT_COLUMNS);
M = insertShape(M, 'line', Coords, 'Color', 'white');  %outputs RGB
M = M(:,:,1) ~= 0;  %convert RGB to binary

imshow(M)