Help with imwarp and affine2d

Hello,

I have been trying to get imwarp to do a simple translation on an image but have been having some problems getting it to work. I have a reference image which I am using to rectify another image. I have the coordinate translations as x=-64 and y=-302. I have set up a coordinate transformation matix but the resulting image does not change.

figure(4);

xform=[1 0 0;0 1 0;64 302 1]; input_image='DSC08069.JPG';

f1=im2double(imread(input_image));

tform=affine2d(xform);

corr_RGBIM=imwarp(f1,tform);

imshow(corr_RGBIM);

title('RGB registered');

Any suggestions?

Answer 1

[David Young]

The problem may be that imwarp() crops the output image to the transformed positions of the corners of the input image. A simple translation does not then have any effect. Try using imwarp_same(), attached, which applies the transformation and crops to the original positions of the corners.

[thiirane]

Thank you David. That did the trick! I appreciate it.

[Daoyu Li]

nice work！

Answer 2

[Image Analyst]

Use the displacement field input argument of imwarp().

Try this to shift the image in x:

fontSize = 20;

grayImage = imread('cameraman.tif');
[rows, columns, numberOfColorChannels] = size(grayImage);
subplot(1, 2, 1);
imshow(grayImage);
axis on;
title('Original Image', 'fontSize', fontSize);
[rows, columns, numberOfColorChannels] = size(grayImage)
% Enlarge figure to full screen.
set(gcf, 'Units', 'Normalized', 'OuterPosition', [0 0 1 1]);

deltaX = 50; % Shift x by 50 pixels.
D = zeros(rows, columns,2);
D(:,:,1) = -deltaX; % Shift x by 50 pixels.
warpedImage = imwarp(grayImage, D);
subplot(1, 2, 2);
imshow(warpedImage);
axis on;
title('Shifted Image', 'fontSize', fontSize);

Try this to shift the image in x and y using the values of -64 and -302 like you specified:

fontSize = 20;

grayImage = imread('concordorthophoto.png');
[rows, columns, numberOfColorChannels] = size(grayImage);
subplot(1, 2, 1);
imshow(grayImage);
axis on;
title('Original Image', 'fontSize', fontSize);
[rows, columns, numberOfColorChannels] = size(grayImage)
% Enlarge figure to full screen.
set(gcf, 'Units', 'Normalized', 'OuterPosition', [0 0 1 1]);

deltaX = 64; % Shift x by 64 pixels.
deltaY = 302; % Shift y by 302pixels.
D = zeros(rows, columns, 2);
D(:,:,1) = -deltaX; % Shift x by 64 pixels.
D(:,:,2) = -deltaY; % Shift x by 302 pixels.
warpedImage = imwarp(grayImage, D);
subplot(1, 2, 2);
imshow(warpedImage);
axis on;
title('Shifted Image', 'fontSize', fontSize);

[DaltonTech]

Hi all,

I am trying to shift an image with imwarp using the displacement option. However I would like to shift the image inside an array slightly bigger than the image, without cropping the image during shifting. Below is a modified version of Image Analyst's code to demonstrate my problem. First I pad grayImage with mean grey level values, then I pad the displacement vector with zeros by the same amount. The subplot shows the result, where unfortunately the right hand side of the image is missing.

clear all
fontSize = 20;
grayImage = imread('cameraman.tif');
imshow(grayImage)
[rows, columns, numberOfColorChannels] = size(grayImage);

subplot(2, 2, 1);
imshow(grayImage);
axis on;
title('Original Image', 'fontSize', fontSize);
% Enlarge figure to full screen.
set(gcf, 'Units', 'Normalized', 'OuterPosition', [0 0 1 1]);

% Displacement matrix
deltaX = 50; % Shift x by 50 pixels.
D = zeros(rows, columns,2);
D(:,:,1) = -deltaX; % Shift x by 50 pixels.

% Warp the original image
warpedImage = imwarp(grayImage, D);
subplot(2, 2, 2);
imshow(warpedImage);
axis on;
title('Shifted Image', 'fontSize', fontSize);

%--------------------------------------------------------------------------
% Additions
%--------------------------------------------------------------------------

% Average grey level of image
meanGreyLevel =  round(mean(grayImage(:)));

% Amount of pixels to pad image
colsPad = 20;
rowsPad = 10;

% Pad option
colPadOption = 'post';
rowPadOption = 'both';

% Pad the columns
tempImage = padarray(grayImage, [0 colsPad], meanGreyLevel, colPadOption);
% Pad the rows
grayImagePadded = padarray(tempImage, [rowsPad 0], meanGreyLevel, rowPadOption);

% Show the image
subplot(2, 2, 3);
imshow(grayImagePadded);
axis on;
title('Padded Image', 'fontSize', fontSize);

% Pad the displacement matrix with 0's to match the padded image
% Pad the columns
tempDisp = padarray(D, [0 colsPad], 0.0, colPadOption);
% Pad the rows
DPadded = padarray(tempDisp, [rowsPad 0], 0.0, rowPadOption);

% Warp the padded image (fill displaced space with avg. grey level)
warpedPaddedImage = imwarp(grayImagePadded, DPadded, 'FillValues', meanGreyLevel);
subplot(2, 2, 4);
imshow(warpedPaddedImage);
axis on;
title('Shifted Padded Image', 'fontSize', fontSize);

Thanks in advance.

[Image Analyst]

They now have an imtranslate() function that's much easier to use than imwarp().

[DaltonTech]

imtranslate looks nice, but 'translation' needs to be a 2-element vector. I need to translate the image using a displacement matrix because in my real problem I need to displace each pixel a different amount. Perhaps I need to modify imwarp to update the world limits so that the outer bounds of the output image are input+displacement value.

Answer 3

[k p]

how to apply affine transform with euclidean distance to find shift between two image after finding matching point pair over image

Answer 4

[Anas Musah]

hello image analyst, I have tried ur code but it gives me this error

    Not enough input arguments.

    Error in imwarp (line 11)
    cornerPoints = round(cornerPoints);

   Error in untitled5 (line 41)
   warpedImage = imwarp(grayImage, D);

pls any help