imrotate won't work in this scenario,any other suggestions?

I want to rotate an image at a pivot point(eg:at the center of the 1st feature, which is NOT the center of the image) without changing the size and shape of the image. I tried imrotate, but it won't work in this scenario. My intention is to stack similar features(eg:features 1 & 2) for each set.How to stack these similar features though? Your response on this regard is appreciated.

[Walter Roberson]

Stack on the display, or stack into a new array? Rotating on the display is easier.

[Chathu]

Thanks alot for your response Walter.

Can you kindly tell me what you meant by 'stack on the display and rotate'? In order to rotate which function should i use? Are you referring to a transformation here (eg:rigid transformation)?

[John BG]

this works, run example_imrotatex.m wuth imrotatex.m and the image in same folder.

[Chathu]

Great! Thank you so much.

Answer 1

[John BG]

Chathu

there is a function in

http://uk.mathworks.com/matlabcentral/fileexchange/57492-rotate-image-around-arbitrary-point

that does the same as the following script, basically

1.- introduce rotation point

2.- 1st padarray

3.- 2nd padarray

4.- imrotate with circle showing furthest pixel radius

the script:

clc;clf;
% clear all
A=imread('countryside.jpg');
figure(1);imshow(A);

angle1=randi([1 359],1,1);

if angle<1
    disp('choose angle > 1º');B4=A;
    return; end;

% measure image size
[sy sx sz]=size(A);

% find image centre [xc yc]
xc=floor(sx/2);
yc=floor(sy/2);

% choose the point [x1 y1] to behave as Z rotation axis
[x1 y1]=ginput(1);

% visualize both [xc yc] and [x1 y1]
hold on;figure(1);plot(x1,y1,'r*','MarkerSize',10);
hold on;figure(1);plot(xc,yc,'ro','MarkerSize',10);

% make sure x1<>xc and y1<>yc
if (x1==xc)
    x1=x1+2*(randi([0 1],1,1)-.5); end;
if (y1==yc)
   y1=y1+2*(randi([0 1],1,1)-.5); end;
ks=0;
if (x1<xc & y1<yc) ks=1; end;
if (x1>xc & y1<yc) ks=2; end;
if (x1>xc & y1>yc) ks=3; end;
if (x1<xc & y1>yc) ks=4; end;

t=0:2*pi/100:2*pi;

switch ks
    case 1 % x1<xc & y1<yc

          D=norm([x1 y1]-[sx sy]);

% 1st pair of padding margins
dx=D+x1-sx;

alpha=atand((sx-x1)/(sy-y1));
s=(2*D^2-2*D^2*cosd(alpha))^.5;
dy=(s^2-(sx-x1)^2)^.5;
% dy=D+y1

dx=floor(dx);dy=floor(dy);

B=padarray(A,[dy,dx],'both');
figure(2);imshow(B);
[sby sbx sbz]=size(B);

hold on;figure(2);plot(x1+dx,y1+dy,'r*','MarkerSize',10);
hold on;figure(2);plot(xc+dx,yc+dy,'ro','MarkerSize',10);

% the origin has been displaced [-dx -dy]
% the point [x1 y1] to become rotation Z axis now has coordinates [dx+x1 +dy+y1]
% visualizing the outer rotation circle
x0d=D*cos(t); xd=x0d+dx+x1;
y0d=D*sin(t); yd=y0d+dy+y1;
hold all;figure(2);plot(xd,yd,'r','LineWidth',1.5);

% visualizing axes after 2nd padarray and origin displacement
hold all;figure(2);plot([1:1:sbx],ones(1,sbx).*(dy+y1),'w','LineWidth',1.5);  % new x axis
hold all;figure(2);plot(ones(1,sby).*(dx+x1),[1:1:sby],'w','LineWidth',1.5);  % new y axis

% calculating 2nd padarray margins
dx2=2*D-sbx;dx2=floor(dx2);
dy2=2*D-sby;dy2=floor(dy2);

B3=padarray(B,[dy2 dx2],50,'pre');
[sb3y sb3x sb3z]=size(B3);
figure(4);imshow(B3);

% visualizing initial image centre [xc yc] and and chosen rotation point [x1 y1]
hold all;figure(4);plot(x1+dx+dx2,y1+dy+dy2,'r*','MarkerSize',10);
hold all;figure(4);plot(xc+dx+dx2,yc+dy+dy2,'ro','MarkerSize',10);

% visualizing outer circle after 2nd padarray
xd2=xd+dx2;
yd2=yd+dy2;
hold all;figure(4);plot(xd2,yd2,'r','LineWidth',2.5);

% visualizing axes after 2nd padarray and origin displacement
hold all;figure(4);plot([1:1:sb3x],ones(1,sb3x).*(dy+y1+dy2),'w','LineWidth',1.5);  % new x axis
hold all;figure(4);plot(ones(1,sb3y).*(dx+x1+dx2),[1:1:sb3y],'w','LineWidth',1.5);  % new y axis

      case 2 % x1>cx & y1<yx

     D=norm([x1 y1]-[0 sy]);

% 1st pair of padding margins
dx=D-x1;

alpha=atand((sy-y1)/(sx-x1));
s=(2*D^2-2*D^2*cosd(alpha))^.5;
dy=(s^2-(sx-x1)^2)^.5;
dy=D-y1;

dx=floor(dx);dy=floor(dy);

B=padarray(A,[dy,dx],'pre');
figure(2);imshow(B);
[sby sbx sbz]=size(B);

hold on;figure(2);plot(x1+dx,y1+dy,'r*','MarkerSize',10);
hold on;figure(2);plot(xc+dx,yc+dy,'ro','MarkerSize',10);

% the origin has been displaced [-dx -dy]
% the point [x1 y1] to become rotation Z axis now has coordinates [dx+x1 +dy+y1]
% visualizing the outer rotation circle
x0d=D*cos(t); xd=x0d+dx+x1;
y0d=D*sin(t); yd=y0d+dy+y1;
hold all;figure(2);plot(xd,yd,'r','LineWidth',1.5);

% visualizing axes after 2nd padarray and origin displacement
hold all;figure(2);plot([1:1:sbx],ones(1,sbx).*(dy+y1),'w','LineWidth',1.5);  % new x axis
hold all;figure(2);plot(ones(1,sby).*(dx+x1),[1:1:sby],'w','LineWidth',1.5);  % new y axis

% calculating 2nd padarray margins
dx2=D-(sx-x1);dx2=floor(dx2);
dy2=D-(sy-y1);dy2=floor(dy2);

B3=padarray(B,[dy2 dx2],50,'post');

[sb3y sb3x sb3z]=size(B3);
figure(4);imshow(B3);

xc2=floor(sb3x/2);yc2=floor(sb3y/2);

% visualizing initial image centre [xc yc] and and chosen rotation point [x1 y1]
hold all;figure(4);plot(xc2,yc2,'r*','MarkerSize',10);
hold all;figure(4);plot(xc2-(x1-xc),yc2-(y1-yc),'ro','MarkerSize',10);

% visualizing outer circle after 2nd padarray
 xd2=xd+xc2;
 yd2=yd+yc2;
hold all;figure(4);plot(xd,yd,'r','LineWidth',2.5);

% visualizing axes after 2nd padarray and origin displacement
hold all;figure(4);plot([1:1:sb3x],ones(1,sb3x).*yc2,'w','LineWidth',1.5);  % new x axis
hold all;figure(4);plot(ones(1,sb3y).*xc2,[1:1:sb3y],'w','LineWidth',1.5);  % new y axis   

      case 3 % x1>xc & y1>yc

    D=norm([x1 y1]-[0 0]);

% 1st pair of padding margins
dx=D-x1;

dy=D-y1;

dx=floor(dx);dy=floor(dy);

B=padarray(A,[dy dx],'pre');
figure(2);imshow(B);
[sby sbx sbz]=size(B);

hold on;figure(2);plot(x1+dx,y1+dy,'r*','MarkerSize',10);
hold on;figure(2);plot(xc+dx,yc+dy,'ro','MarkerSize',10);

x0d=D*cos(t); xd=x0d+dx+x1;
y0d=D*sin(t); yd=y0d+dy+y1;
hold all;figure(2);plot(xd,yd,'r','LineWidth',1.5);

% visualizing axes after 2nd padarray and origin displacement
hold all;figure(2);plot([1:1:sbx],ones(1,sbx).*(dy+y1),'w','LineWidth',1.5);  % new x axis
hold all;figure(2);plot(ones(1,sby).*(dx+x1),[1:1:sby],'w','LineWidth',1.5); % new y axis

% calculating 2nd padarray margins
dx2=2*D-sbx;dx2=floor(dx2);
dy2=2*D-sby;dy2=floor(dy2);

B2=padarray(B,[0 dx2],50,'post');
figure(3);imshow(B2);
B3=padarray(B2,[dy2 0],100,'post');
[sb3y sb3x sb3z]=size(B3)
figure(4);imshow(B3);

xc2=floor(sb3x/2);yc2=floor(sb3y/2);

% visualizing initial image centre [xc yc] and and chosen rotation point [x1 y1]
hold all;figure(4);plot(xc2,yc2,'r*','MarkerSize',10)'
hold all;figure(4);plot(xc2-(x1-xc),yc2-(y1-yc),'ro','MarkerSize',10)'

% visualizing outer circle after 2nd padarray
xd2=x0d+xc2;
yd2=y0d+yc2'
hold all;figure(4);plot(xd2,yd2,'r','LineWidth',2.5)'

% visualizing axes after 2nd padarray and origin displacement
hold all;figure(4);plot([1:1:sb3x],ones(1,sb3x).*yc2,'w','LineWidth',1.5);  % new x axis
hold all;figure(4);plot(ones(1,sb3y)*xc2,[1:1:sb3y],'w','LineWidth',1.5);  % new y axis      

      case 4 % x1<xc & y1>yc

% D is distance between [x1 y1] and [xc yc], also D=pdist2([x1 y1],[sx 0])
D=norm([x1 y1]-[sx 0])

% 1st pair of padding margins
dx=D+x1-sx

alpha=atand((sx-x1)/y1)
s=(2*D^2-2*D^2*cosd(alpha))^.5
dy=(s^2-(sx-x1)^2)^.5
% dy=D-y1

dx=floor(dx);dy=floor(dy);

B=padarray(A,[dy,dx],'both');
figure(2);imshow(B)
[sby sbx sbz]=size(B)

hold on;figure(2);plot(x1+dx,y1+dy,'r*','MarkerSize',10)
hold on;figure(2);plot(xc+dx,yc+dy,'ro','MarkerSize',10)

% the origin has been displaced [-dx -dy]
% the point [x1 y1] to become rotation Z axis now has coordinates [dx+x1 +dy+y1]
% visualizing the outer rotation circle

x0d=D*cos(t); xd=x0d+dx+x1;
y0d=D*sin(t); yd=y0d+dy+y1;
hold all;figure(2);plot(xd,yd,'r','LineWidth',1.5);

% visualizing axes after 2nd padarray and origin displacement
hold all;figure(2);plot([1:1:sbx],ones(1,sbx).*(dy+y1),'w','LineWidth',1.5)  % new x axis
hold all;figure(2);plot(ones(1,sby).*(dx+x1),[1:1:sby],'w','LineWidth',1.5)  % new y axis

% calculating 2nd padarray margins
dx2=2*D-sbx;dx2=floor(dx2);
dy2=2*D-sby;dy2=floor(dy2);

B2=padarray(B,[0 dx2],50,'pre');
figure(3);imshow(B2);
B3=padarray(B2,[dy2 0],100,'post');
[sb3y sb3x sb3z]=size(B3)
figure(4);imshow(B3);

% visualizing initial image centre [xc yc] and and chosen rotation point [x1 y1]
hold all;figure(4);plot(x1+dx+dx2,y1+dy,'r*','MarkerSize',10)
hold all;figure(4);plot(xc+dx+dx2,yc+dy,'ro','MarkerSize',10)

% visualizing outer circle after 2nd padarray
xd2=xd+dx2
yd2=yd
hold all;figure(4);plot(xd2,yd2,'r','LineWidth',2.5)

% visualizing axes after 2nd padarray and origin displacement
hold all;figure(4);plot([1:1:sb3x],ones(1,sb3x).*(dy+y1),'w','LineWidth',1.5)  % new x axis
hold all;figure(4);plot(ones(1,sb3y).*(dx+x1+dx2),[1:1:sb3y],'w','LineWidth',1.5)  % new y axis

      otherwise
          disp('\nerror\n')
          return
  end
  % 
  % imsave(B3)
  imwrite(B3,'padded_image.jpg')
  % padded image: B3
  % size square containing padded image: [sb3x sby3]
  % center padded image: [xc2 yc2]
  % points outer circle centre before rotation [xd2 yd2]
  % radius: D
B4=imrotate(B3,angle);
[sb4y sb4x sb4z]=size(B4);
figure(5);imshow(B4);

% visualizing initial image centre [xc yc] and and chosen rotation point [x1 y1]
hold all;figure(5);plot(floor(sb4x/2),floor(sb4y/2),'r*','MarkerSize',15);
% hold all;figure(5);plot(xc+dx+dx2,yc+dy,'ro','MarkerSize',10);

% visualizing outer circle after 2nd padarray
xd4=x0d+floor(sb4x/2);
yd4=y0d+floor(sb4y/2);
hold all;figure(5);plot(xd4,yd4,'r','LineWidth',2.5);
if angle>0 str1='CCW'; end;
if angle<0 str1='CW'; end;
figure(5);legend(['angle : ' num2str(angle) ' º ' str1],['Radius : ' num2str(floor(D)) ' pixels'],'Location','Northwest');

imwrite(B4,'im_padded_rotated.jpg');

If you find this answer of any help solving your question,

please click on the thumbs-up vote link, or mark is accepted answer

thanks in advance

John

[e-mail]

[Chathu]

Hi John, sorry for the late response.This code looks like an excellent code. Let me apply it to my scenario and get back to you soon. Thank you sooo much for your kindest effort. Really appreciate it.

[John BG]

I just received an email from your address mentioning an error. Would you please explain the error, how can i reproduce it, so i can fix it?

John

[e-mail]

[Chathu]

angle1=randi([1 359],1,1);

angle1 should be corrected as angle.

That's it and it works really well. Thank you so much John. Really appreciate your effort.

Answer 2

[Walter Roberson]

Parent each image within its own hgtransform() . You can then construct appropriate rotation and translation matrices to rotate each of them around arbitrary points and move them around on the display.

This is a solution for the case where what you care about is what shows up on the display.

[Chathu]

Thank you so much for your response, Walter. That helps.

But i was wondering whether i could use 'imtransform' function to perform the image for a Rigid Transformation?

[Walter Roberson]

Your requirement that the size and shape of the image not be altered makes it difficult to use imtransform or imwarp.

Consider, for example, a 400 x 800 grayscale image and a simple 90 degree rotation: your requirement is that the output would still be 400 x 800 but contain the rotated image. You can see that in order to do that, you would have to do the equivalent of

rotated_image = [zeros(400,300), rot90( imresize(Original_Image, [200 400]) ), zeros(400,300)]

just like the black bars that show up in http://www.mathworks.com/help/images/ref/imtransform.html#zmw57dd0e85423

This is likely to interfere with your attempts to match up anything other than the center point that you are rotating the image around.

[Chathu]

Walter, thank you so much for your solution. I know, pivot point may be an issue as it is not the image's center.

Answer 3

[Image Analyst]

Exactly what does "without changing the size and shape of the image" mean to you? And why is that necessary?

When you rotate an image, say a rectangle, it will become a rotated rectangle, like a diamond shape. However since images can't be diamond shapes, it will pad the "blank" space that got uncovered with zeros. The corners rotated up and down will now make the image have more rows than before. So you can either enlarge the canvass, or clip/crop the canvass to make it have the same number of rows and columns as the original. imrotate() can either enlarge the canvass or crop it, depending on what you pass in for the input.

We don't know what you mean. Please clarify. To rotate about some point that is not at the center of the image, you might want to use padarray() to enlarge the canvass first so that the center or rotation is the center of the new, enlarged canvass. That's the easiest way though not the most memory efficient.

[Chathu]

Image Analyst, thanks alot for your response. I was referring to Rigid transformation. The reason is i want to register the images by keeping the image shape & size constant. Well,i want to rotate the image by the center of the 1st feature(kindly note that all features what i meant here are circles) which is not the center of the whole image. I'm sorry if i confuse you.Anyhow really appreciate your response though.

[Image Analyst]

You can see from all the responses here just how confusing your explanations are. You might make things a LOT clearer by just posting your image and saying what "feature" you want to rotate around and why. And if there are, say, 5 features in the image, which do you consider to be the first? And what is a "feature" to you? To most people, a feature is something like the area, perimeter, intensity, centroid or some other measurement of an object region. Like if you have a photo of nuts and bolts, the "objects" would be the nuts and bolts and the "features" would be the area and brightness and Euler number. But it kind of sounds like you may be calling the whole object a "feature" which is contrary to standard nomenclature. Rotating around the first feature, say, area, would not make sense. Rotating around the first bolt would make sense but you'd have to define which bolt is the "first" bolt. Again, post a picture.

[Chathu]

Image Analyst, i know i should have post it before. I am sorry. But here you go.