'image' behaviour for 'imwrite'

I have these lines of code to display values of a matrix as an image:

x=linspace(-20, 20, 1000);
y=linspace(0, 115, 1482);
image(x, y, img)
colormap(gray(127))
axis('image')
axis([-20 20 30 90])

img is a 1482x1000 matrix. The specific numers don't matter. The code is from here if that's helpfull.

This code creates a figure with the correctly scaled and framed image. My goal now is not to open a figure, but to save the image to a file (preferably .png). If I just use

imwrite(img, save_path)

then the aspect ratio of the image is wrong.

So my question is basically: how can I use the syntax or behaviour of 'image' and don't open a figure but save the image. I don't want to save the figure that is created by 'image'. The resolution is lower and there is a white border.

Answer 1

[Image Analyst]

If you don't want to save img (perhaps it is already saved) but would rather save a screenshot, then you can use exportgraphics(). I presume you have a 2020 release but you blew by the field on the post submission and forgot to tell us your release. exportgraphics() came out this year. Otherwise try saveas().

[Seba]

I have a 2020 release (sorry, forgot to mention it), but my goal is to save the image itself, just like it would look if I'd open it as a figure. I create may of those matrices in a loop, so opening a figure every time wouldn't be that great as well.

[Image Analyst]

And so did you not like exportgraphics()? It's basically a screenshot. What happened when you tried it?

If you want the original size img (instead of whatever size it happens to be displayed at) then you'd have to apply your colormap with ind2rgb() since you have only 127 levels instead of the normal 256. And then call rgb2gray() to get it back to grayscale

grayImage = uint8(rgb2gray(ind2rgb(img, gray(127))));
imwrite(grayImage, fileName);

[Seba]

yes, I need the original size image. The color is not the problem, the problem is that the aspect ratio isn't correct because I can't use the arrays x and y in imwrite.

[Image Analyst]

iwrite just makes an image where each element is a square pixel. If you need a different aspect ratio, like one to match your x and y, then you need to computer the new size [rows, columns] based on the range of x and y, then call imresize().

How big is the original img?

[rows, columns, numberOfColorChannels] = size(img)  % No semicolon

Now, you have

x=linspace(-20, 20, 1000);
y=linspace(0, 115, 1482);

so x is 1000 columns wide, and has values that range from -20 to +20, and y is 1482 rows tall and has values that range from 0 to 115.

Tell me how many rows tall and columns wide you'd want your output image? 1482 tall and 1000 wide? Or do you want to use the aspect ratio of x and y and one of their dimensions to compute the size? Like

% Compute the aspect ratio of width over height:
aspectRatio = (x(end)-x(1)) / (y(end)-y(1));
% Use that factor to scale the original image's rows and columns.
outputRows = rows
outputColumns = round(columns * aspectRatio)
outputImage = imresize(img, [outputRows, outputColumns]);
% outputImage will have the same number of rows as img but the aspect ratio of x and y

Alternatively you could have the original number of columns of img:

% Compute the aspect ratio of width over height:
aspectRatio = (x(end)-x(1)) / (y(end)-y(1));
% Use that factor to scale the original image's rows and columns.
outputRows = round(rows / aspectRatio)
outputColumns = columns
outputImage = imresize(img, [outputRows, outputColumns]);
% outputImage will have the same number of rows as img but the aspect ratio of x and y

Or you could have the same number of rows as y has elements

% Compute the aspect ratio of width over height:
aspectRatio = (x(end)-x(1)) / (y(end)-y(1));
% Use that factor to scale the original image's rows and columns.
outputRows = round(length(y) / aspectRatio)
outputColumns = length(x)
outputImage = imresize(img, [outputRows, outputColumns]);
% outputImage will have the same number of rows as img but the aspect ratio of x and y

The aspect ratio of x and y is 40/115 = 0.35, while the aspect ratio of the number of elements of them is 1000/1482 = 0.67. Why are these not the same? An image has to have square pixels (unless you plan on doing tons of math to account for non-square pixels).

There are a lot of possible combinations and I don't want to list them all. Why don't you just tell me how many pixels you expect the output image to be?