How to prevent imshow() from resizing the figure?

I'm still at the stage of figuring out, to the best of my ability and according to the information I have access to, the interactions between a figure and commands like imshow(). I have a script where I first create an 800x1000 pixels figure but as soon as the imshow() executes, the figure is automatically resized, why? And how do I keep control over the figure (window) size? Here's my code :

f = figure('Position', [100 100 800 1000],  'Name','Roger Breton');

% imread resizes the figure?

img=imread('Fleur_3x4.jpg');

I've been searching the document to no avail :(

Answer 1

[Roger Breton]

Here's the final code :

global img;

f = figure('color', 'w', 'MenuBar', 'none', 'Position', [500 500 800 600])

global ax;
ax = axes

img = imread('Fleur_3x4.jpg');

imshow(img);

TailleImg = size (img);  

global ah;
ah = annotation(f, 'line','Units', 'Pixels', 'Position', [311 183 TailleImg(2) 0], 'Color','k','LineWidth',10);

hold on
set(f, 'Position', [10 10 800 600]);
set(f, 'Resize','off');
set(ax, 'Units', 'Pixels');
set(ax, 'OuterPosition', [1 1 800 600]);
set(ax, 'InnerPosition', [311 191 TailleImg(2) TailleImg(1)]);
set(ax, 'Position', [311 191 TailleImg(2) TailleImg(1)]);;

global Inner;
[Inner] = get(ax, 'InnerPosition');

set (gcf, 'WindowButtonMotionFcn', @mouseMove);

function mouseMove (object, eventdata)
global img;
global ah;
global ax;
global Inner;

CP = round(get (ax, 'CurrentPoint'));
x  = CP(1,1)
y  = CP(1,2)

IntX = cast(x, 'uint16')
IntY = cast(y, 'uint16')


if IntX < 1 
    ah.Color=[1, 1, 1];
    xlabel(ax, ' ');
    return;
end

if IntX > Inner(3) 
     ah.Color=[1, 1, 1];
        xlabel(ax, ' ');
   return;
end

if IntY < 1 
     ah.Color=[1, 1, 1];
        xlabel(ax, ' ');
    return;
end

if IntY > Inner(4) 
     ah.Color=[1, 1, 1];
        xlabel(ax, ' ');
   return;
end

RGB = img(IntY,IntX, :);

Lab = rgb2lab(RGB)

rouge = double(RGB(1,1,1))/ 255.00
vert = double(RGB(1,1,2))/ 255.00
bleu = double(RGB(1,1,3))/ 255.00

ah.Color=[rouge, vert, bleu]; 

Temp1 = ['(X,Y) = ', num2str(IntX), ', ', num2str(IntY)];
Temp2 = ['RGB = ', num2str(RGB(1,1,1)), ', ', num2str(RGB(1,1,2)), ' , ', num2str(RGB(1,1,3))];
Temp3 = ['Lab = ', num2str(Lab(1,1,1),4), ', ', num2str(Lab(1,1,2),4), ' , ', num2str(Lab(1,1,3),4)];
xlabel(ax, [{Temp1, Temp2, Temp3}]);

end

I'm sure there's plenty of room for lot's of optimizations but, at the stage I'm at, learning Matlab, I feel like putting my feet up. The code would deserve adequate documentation, I know... I still have a ton of work to do, like, this works with one 'single image' but it's a 'proof of concept' for me. At least, I know what to expect now and the results meet my initial goals.

Answer 2

[Walter Roberson]

h2 = imshow(img, 'InitialMagnification', 'fit')

will examine the axes size and will set the image to just fill the axes drawing area.

If you want control over your axes and figure, I recommend against using imshow().

imshow() first checks for a current axes, and then compares the axes position to the default axes position within the figure. If they are the same size, then imshow() deletes the axes and creates a new one -- destroying any properties of the axes. If I recall correctly, it does not do this if "hold" is on, though.

Also, if you pass XData and YData to imshow() then it fits the image into those coordinates, same as if you had used image() or imagesc() with the same XData and YData. I would need to review the code to see whether it still deletes the axes in that situation.

image() on the other hand, never deletes the current axes, and never resizes the current figure. It is more robust when you want control over sizes and axes properties.

[Walter Roberson]

I'm still not clear on the relationship between axes, figure and images...

In MATLAB, a figure() is a drawing window that can have behaviour attached to it. Controls and menu items can be attached. A MATLAB figure() is the same thing as a GUI.

A figure() is a container object. Historically it was the outer-most container object other than the graphics root, but there turns out to be a layer above that (which I would have to research as I keep forgetting how they work.)

A figure can contain several controls, several panels (uipanels), several tabs, several axes. Panels and tabs can contain axes as well. Panels and tabs can be nested. Axes cannot be nested.

An axes has a single coordinate system for each of x, y, z (or for appropriate dimension names for polar axes). You cannot draw with different coordinate systems within a single axes. You can place several axes at the same position in the figure (or sub-container)

An image is one particular graphics object that can be placed within an axes; there are also lines, surfaces, patches, text, and several increasingly specialized objects.

[Walter Roberson]

I should perhaps add that inside axes(), the coordinate system for drawing purposes is primarily "data" coordinates. Outside of axes, the coordinates are usually either normalized coordinates or pixels.

[Roger Breton]

Thank you SO VERY MUCH for the kind help, Walter. I've been experimenting quite a bit on my own and this is what I discovered. Here's one exercise that satisfies my requirements for now (which is to understand what's going on under the hood and how to get control over Matlab drawing (without any pretention), if at all possible) :

global img;

% La Taille de la figure exclut la barre de titre
f = figure('color', 'w', 'MenuBar', 'none', 'Position', [500 500 800 600])

% fh = figure
ax = axes
% allAxesInFigure = findall(fh,'type','axes')

% OuterPosition = Taille de Figure
ah(1) = annotation('rectangle',ax.OuterPosition,'Color', 'r','LineWidth',2);
ah(2) = annotation('rectangle',ax.InnerPosition,'Color','g','LineWidth',6);
ah(3) = annotation('rectangle',ax.Position,'Color','b','LineWidth',2);

% To get around this, call "hold off" after you call axes() and before you call imshow().


img = imread('Fleur_3x4.jpg');
imshow(img);

hold on

set(f, 'Position', [500 500 800 600]);

set(ax, 'Units', 'Pixels');
set(ax, 'OuterPosition', [1 1 800 600]);
set(ax, 'InnerPosition', [311 191 216 293]);
set(ax, 'Position', [311 191 216 293]);

% Ne change pas le OuterPosition ...set(ax, 'OuterPosition', [105 67 620 489]);
% allAxesInFigure = findall(fh,'type','axes');

dummy = 3;

set (gcf, 'WindowButtonMotionFcn', @mouseMove);

function mouseMove (object, eventdata)
global img;

C = round(get (gca, 'CurrentPoint'));
PetitX = num2str(C(1,1))
PetitY = num2str(C(1,2))

RGB = img(PetitX, PetitY, :)
Lab = rgb2lab(RGB);

% title(gca, ['(X,Y) = (', num2str(C(1,1)), ', ',num2str(C(1,2)), ')']);
% title(gca, [RGB(1), RGB(2), RGB(3)]);
% TitleH = title(gca, ['(X,Y) = (', PetitX, ', ', PetitY, ' RGB = ',...
%     num2str(RGB(1)), ', ', num2str(RGB(2)), ' ,', num2str(RGB(3)), ') Lab = ',...
%     num2str(Lab(1)), ', ', num2str(Lab(2)), ' ,', num2str(Lab(3)), ')']);

Temp1 = ['(X,Y) = ', PetitX, ', ', PetitY];
Temp2 = ['RGB = ', num2str(RGB(1)), ', ', num2str(RGB(2)), ' , ', num2str(RGB(3))];
Temp3 = ['Lab = ', num2str(Lab(1),4), ', ', num2str(Lab(2),4), ' , ', num2str(Lab(3),4)];

xlabel(gca, [{Temp1, Temp2, Temp3}]);
%xlabel(gca, [Temp1, Temp2, Temp3]);

% xlabel(gca, ['(X,Y) = (', PetitX, ', ', PetitY, ') RGB = (',...
%     num2str(RGB(1)), ', ', num2str(RGB(2)), ' ,', num2str(RGB(3)), ') Lab = ',...
%     num2str(Lab(1),2), ', ', num2str(Lab(2),2), ' ,', num2str(Lab(3),2), ')']);

% ax = gca;
% ax.TitleVerticalAlignment ='bottom';
% ax.TitleHorizontalAlignment = 'left';

% set(TitleH, 'Position', [10, 10], ...
%   'VerticalAlignment', 'bottom', ...
%   'HorizontalAlignment', 'center')

% set(TitleH, 'VerticalAlignment', 'bottom', 'HorizontalAlignment', 'center');

end

% 
% set(handles.axes7,'Units','pixels');
% resizePos = get(handles.axes7,'Position');
% myImage= imresize(myImage, [resizePos(3) resizePos(3)]);
% axes(handles.axes7);
% imshow(myImage);
% set(handles.axes7,'Units','normalized');

% allAxesInFigure = findall(figureHandle,'type','axes');
% isAxes = strcmp('axes',get(listOfHandles,'type')); true or false

I purposely left all the comments in, to show you the various 'directions' I tried to take while "learning" how to tame this 'monster'... This is what I'm getting now :

I'll sure give your code a try, and report ... :-)

[Roger Breton]

I tried and compared the tree following ways to initialise imshow() :

img = imread('Fleur_3x4.jpg');

h2 = imshow(img, 'InitialMagnification', 'fit')

h2 = imshow(img, 'InitialMagnification', 100) % default

h2 = imshow(img, 'XData',[1 216], 'YData', [1 293]) % image is 216x293 pixels

In all three cases, the image is automatically resized when the window (figure) is resized.

Initially, I aim to keep the size of the image not to scale with change in the figure resize.

See, one of the frustration I stumbled upon, was the fact that my third label line was getting "chopped" at the bottoom and I could not find any way around that :

To some extent, I don't really care to control the axes and that the image is resized but I do care about my stupid CIE Lab values not being completely visible in the figure :(

Ideall, I'd have to do is to throw in an impixinfo() command with an option of Lab values (or both RGB and Lab) values and I'd be a happy camper...

[Roger Breton]

I just tried image() and all my three lines of labels made intact into the figure!!!

The picture is "distorted", e.g. 'scaled horizontally'.

Maybe if I specify the figure size and axes before calling image()?

[Walter Roberson]

"Parent" graphics as much as you can.

global img;

% La Taille de la figure exclut la barre de titre
f = figure('color', 'w', 'MenuBar', 'none', 'Position', [500 500 800 600])

f = 
  Figure (1) with properties:

      Number: 1
        Name: ''
       Color: [1 1 1]
    Position: [500 500 800 600]
       Units: 'pixels'

  Show all properties

% fh = figure
ax = axes(f)

ax = 
  Axes with properties:

             XLim: [0 1]
             YLim: [0 1]
           XScale: 'linear'
           YScale: 'linear'
    GridLineStyle: '-'
         Position: [0.1300 0.1100 0.7750 0.8150]
            Units: 'normalized'

  Show all properties

% allAxesInFigure = findall(fh,'type','axes')

% OuterPosition = Taille de Figure
ah(1) = annotation(f, 'rectangle',ax.OuterPosition,'Color', 'r','LineWidth',2);
ah(2) = annotation(f, 'rectangle',ax.InnerPosition,'Color','g','LineWidth',6);
ah(3) = annotation(f, 'rectangle',ax.Position,'Color','b','LineWidth',2);

% To get around this, call "hold off" after you call axes() and before you call imshow().


%img = imread('Fleur_3x4.jpg');
img = imread('kobi.png');

imshow(img, 'Parent', ax);

hold(ax, 'on')

set(f, 'Position', [500 500 800 600]);

set(ax, 'Units', 'Pixels');
set(ax, 'OuterPosition', [1 1 800 600]);
set(ax, 'InnerPosition', [311 191 216 293]);
set(ax, 'Position', [311 191 216 293]);

% Ne change pas le OuterPosition ...set(ax, 'OuterPosition', [105 67 620 489]);
% allAxesInFigure = findall(fh,'type','axes');

dummy = 3;

set(f, 'WindowButtonMotionFcn', @mouseMove);

function mouseMove (object, eventdata)
global img;

ax = ancestor(object, 'axes');

C = round(get (ax, 'CurrentPoint'));
PetitX = num2str(C(1,1))
PetitY = num2str(C(1,2))

RGB = img(PetitX, PetitY, :)
Lab = rgb2lab(RGB);

% title(gca, ['(X,Y) = (', num2str(C(1,1)), ', ',num2str(C(1,2)), ')']);
% title(gca, [RGB(1), RGB(2), RGB(3)]);
% TitleH = title(gca, ['(X,Y) = (', PetitX, ', ', PetitY, ' RGB = ',...
%     num2str(RGB(1)), ', ', num2str(RGB(2)), ' ,', num2str(RGB(3)), ') Lab = ',...
%     num2str(Lab(1)), ', ', num2str(Lab(2)), ' ,', num2str(Lab(3)), ')']);

Temp1 = ['(X,Y) = ', PetitX, ', ', PetitY];
Temp2 = ['RGB = ', num2str(RGB(1)), ', ', num2str(RGB(2)), ' , ', num2str(RGB(3))];
Temp3 = ['Lab = ', num2str(Lab(1),4), ', ', num2str(Lab(2),4), ' , ', num2str(Lab(3),4)];

xlabel(ax, [{Temp1, Temp2, Temp3}]);
%xlabel(gca, [Temp1, Temp2, Temp3]);

% xlabel(gca, ['(X,Y) = (', PetitX, ', ', PetitY, ') RGB = (',...
%     num2str(RGB(1)), ', ', num2str(RGB(2)), ' ,', num2str(RGB(3)), ') Lab = ',...
%     num2str(Lab(1),2), ', ', num2str(Lab(2),2), ' ,', num2str(Lab(3),2), ')']);

% ax = gca;
% ax.TitleVerticalAlignment ='bottom';
% ax.TitleHorizontalAlignment = 'left';

% set(TitleH, 'Position', [10, 10], ...
%   'VerticalAlignment', 'bottom', ...
%   'HorizontalAlignment', 'center')

% set(TitleH, 'VerticalAlignment', 'bottom', 'HorizontalAlignment', 'center');

end

% 
% set(handles.axes7,'Units','pixels');
% resizePos = get(handles.axes7,'Position');
% myImage= imresize(myImage, [resizePos(3) resizePos(3)]);
% axes(handles.axes7);
% imshow(myImage);
% set(handles.axes7,'Units','normalized');

% allAxesInFigure = findall(figureHandle,'type','axes');
% isAxes = strcmp('axes',get(listOfHandles,'type')); true or false

[Roger Breton]

Thank you for your help, Walter... I'm not sure I can digest all that code...

I'll give it a try...

I am running into 'trouble' with this line of code :

RGB = img(PetitX, PetitY, :)

Unknowingly, this is returning a 3×4×3 uint8 array? My life would be easier if I could index into the Image array in a way I would only get 1 set of RGB values back? It is like a 12 pixels set? With wildly varying RGB values like this :

3×4×3 uint8 array

RGB(:,:,1) =

229 174 174 210

187 86 86 141

226 235 235 222

RGB(:,:,2) =

25 21 21 55

23 12 12 22

113 49 49 24

RGB(:,:,3) =

62 41 41 59

50 3 3 16

33 54 54 73

I don't see how I could tell Matlab to restricts its 'range'?

[Roger Breton]

I think I found my mistake... First, I modified my test image in Photoshop, to have a "solid" area like this :

See the little "yellow" square with a black border? It is located at x = 179 and y = 143, from the top, left hand side of the image. So I wrote this code, to test the value, right after reading the image in :

img = imread('Fleur_3x4.jpg');

RGB = img(179, 143, :)

See? I thought, naively, that Matlab would expect the argument as "Row", "Column", but that's not it. So I changed the argument order to :

img = imread('Fleur_3x4.jpg');

RGB = img(143, 179, :)

And got this RGB value :

RGB(:,:,1) =

254

RGB(:,:,2) =

199

RGB(:,:,3) =

55

Which seems exactly what I want, one RGB triplet only! I'll give it a try in the rest of my code...

[Walter Roberson]

MATLAB does use row, column. However, x is columns and y is rows.

[Roger Breton]

I think I am having difficulty with 'data types'?

If I use this line of code, to index into the Image array :

RGB = img(143, 179, :)

I get exactly what I expect, a 1×1×3 uint8 array.

But if I used the same code as before :

PetitX = num2str(C(1,1)) % X = 179

PetitY = num2str(C(1,2)) % Y = 143

OR this :

PetitX = uint8(C(1,1)) % X = 179

PetitY = uint8(C(1,2)) % Y = 143

RGB = img(PetitY, PetitX, :)

OR this :

RGB = img(uint8(C(1,2), C(1,1), :)

I invariably get a 3×3×3 uint8 array? The trouble is that PetitX and PetitY are coming out as 1x3 char? Pardon my ignorance but in some other programming languages, the type 'char' means a 'caracter', like an ASCII character. Which could explain why I am getting a 3x3 array instead of just a 1x3 vector?

[Walter Roberson]

https://www.mathworks.com/matlabcentral/answers/62653-convert-currentpoint-to-x-and-y-value#answer_74247 "The CurrentPoint property of the axes replies a 2x3 array, which defines the viewing line through a 3D scene."

when you num2str then you are converting the numbers to character vector, such as 184 decimal to '184' character vector, which is a 1x3 array of character.

When you use a character vector as a subscript, the content of the character vector is not analyzed as a potential representation of a number, so the characters '184' are not treated as numeric 184. Instead, in most cases, the character vector is converted to a vector of the underlying Unicode code points, so '184' would be converted to the numeric vector [49 56 52]. You would then be indexing with a pair of 1x3 vectors. When you index with a non-scalar, the output is the result of indexing with all combinations selected from each index. If the second numeric value had been (say) 73 then your num2str would create the 1x2 character vector '73' and as an index go for the underlying Unicode [55 51] . The indexing by '184','73' would then effectively be the same as indexing by ([49 56 52], [55 51]) which would give you a 3x2 result.

Answer 3

[Roger Breton]

Strictly speaking, I could always issue a set('Position', ....) using my original figure size vector, after executing the imshow() like so :

f = figure('Position', [100 100 800 1000],  'Name','Roger Breton');
h2 = imshow(img);
f.Position = [100 100 800 1000];

But, 1) that seems an overkill? And 2) the picture area is now drawn in proportion to the figure and appears larger.

Not ideal. I bet the solution is to turn the 'fit' option off or something like that...

[Roger Breton]

Well, I tried this code :

h2 = imshow(img);
set(h2, 'InitialMagnification',100);
f.Position = [100 100 800 1000]; 

and I get an error :

Unrecognized property InitialMagnification for class Image.

Back to square one...

[Roger Breton]

Looks like my anwer is to learn about axes...

I'm still not clear on the relationship between axes, figure and images...