how to match plot and image coordinates ??

Dea all, I am doing some image analysis. My basic problem is that images and plots in matlab have different conventions on their coordinate system (CS) and this screws me. this makes some stuff really tricky and i could not yet solve that:

clear all;close all

% load image and view it
I = imread('coins.png');
figure(1); imagesc(I); colormap(gray); axis equal; axis tight; hold on;
% (i take imagesc rather than imshow for the axis ticks)

%find the circles & centers & plot
[centers, radii, metric] = imfindcircles(I,[15 30]);
viscircles(centers, radii,'EdgeColor','b');

% plotting them works well:
plot(centers(:,1),centers(:,2));

% now lets modify the image around the first center
x=round(centers(1,1))-5:round(centers(1,1))+5; %xrange around 1st center
y=round(centers(1,2))-5:round(centers(1,2))+5; %yrange around 1st center

I(x,y)=0; % put its surrounding to zero

figure(2)
imagesc(I); colormap(gray); axis equal; axis tight; hold on;

% => fail!

% the plotting has different coordinates:
plot(x,y,'.')

additionally, I want the image CS to match the plot CS, such that the pixel (0,0) is in the lower left, (1,0) it right neigbor and (0,1) just above (0,0). I manageed to do this via

figure(); imagesc(I'); set(gca,'YDir','normal'); %mind the transpose!

but this screws up coordinate matching completely. can you help me?

In the end I need plot(point1,point1) and I(point1,point2) to be displayed in the same spot, within the standard plot CS.

Thank you very much for any suggestions!

[Matlabber 1.0]

another illustrating example:

I = imread('coins.png');
figure()
[a,b]=size(I);
plot([0,a],[0,b]); hold on
imagesc(I);colormap(gray); axis equal; axis tight;
bx=[1,50]
by=[10,100]
I(bx(1):bx(2),by(1):by(2))=0;
imagesc(I);colormap(gray); axis equal; axis tight;
plot([bx(1),bx(2),bx(2),bx(1)],[by(1),by(1),by(2),by(2)])
% => 2 diferent boxes!

[Image Analyst]

chris, did you see my answer? Evidently not because you're making the same mistake. Again: row, column is NOT x,y!!! It's just NOT.

Answer 1

[Image Analyst]

You have made a very common error that trips up all beginners eventually, and even sometimes experts:

(x,y) are not the coordinates of matrices in MATLAB!

And images are matrices. The coordinates of matrices are (row, column) which is NOT (x,y) - it's (y,x). So when you did this:

I(x,y)=0; % put its surrounding to zero

You should have done this:

I(y, x) = 0; % put its surrounding to zero

[Matlabber 1.0]

yes this is exactly what I found out. it becomes especially annoying when i start using improfile or finding centers of circles, as would your indicate that it is impossible in matlab to get some unified approach for this? or would i rather make up my mind which way i need to use the coortinates every time?

btw sorry for not seeing your anser in the first place, i had the browser window not updated too long :)

[Image Analyst]

No, the x,y / row,column thing has been around as long as there has been digital images, and it's not just MATLAB that will have the issue. It's just a convention, like the convention that plots with cartesian coordinates have the origin at the lower left and y increases as you rise vertically, while for images and arrays, the origin is at the upper left and y (row, line number) increases as you go down vertically. They're just conventions you have to get used to.

You just have to get in the right mindset. Sometimes people like to think in terms of x,y and sometimes they like to think in terms of row, column. You just have to look out. For example if you're using getinput() on an image, it returns x,y. They might have chosen it to return row, column which would work fine for images, but it would be reversed to the way people are used to for regular x,y types of plots. There are a lot of functions in MATLAB that return a 2D coordinate - some return row, column (like find() function) and some return x,y - so you just have to check and be careful.

[Matlabber 1.0]

sounds not very promising. i guess ill have to transpose on the correct occasions, alsways

Answer 2

[Matlabber 1.0]

use the transpose " ' " and set the y-axis to "normal", in order to align the image-coordinate system to that of the internal matrix notation:

im=peaks(100);                           % make an image
im(1:20;20:40)=max(im(:));               % modify a specific region
imshow(im',[]);set(gca,'YDir','normal')  % plot it such that the image coordinate system matches that of the internal matrix notation

so you can do all your image/ data processing as usual. once you plot images (also in combination with normal plots), transpose the image before plotting and make sure the y axis of your plot is set to normal.