Position based image stitching

I have 4 same size tile images and they are partially overlapped (assume 15% each).

I don't want to use stitching algorithms like feature based stitching but just array images in rectangle mold like above image.

Thanks!

Answer 1

[Walter Roberson]

The below uses a facility introduced in R2016a

overlap_fraction = 15/100;

[r, c, p] = size(FirstImage);    %we assume the others are the same size
overlap_c = round(overlap_fraction * c);
new_c = c + c - overlap_c;
overlap_r = round(overlap_fraction * r);
new_r = r + r - overlap_r;

temp_image1 = nan(new_r, new_c, p );   %double
temp_image2 = temp_image1;
temp_image3 = temp_image1;
temp_image4 = temp_image1;

temp_image1(1:r, 1:c, :) = FirstImage;
temp_image2(1:r, end-c+1:end, :) = SecondImage;
temp_image3(end-r+1:end, 1:c, :) = ThirdImage;
temp_image4(end-r+1:end, end-c+1:end, :) = FourthImage;

temp_image = cat(4, temp_image1, temp_image2, temp_image3, temp_image4);

%this line requires R2016a or later
registered_image = cast( mean(temp_image, 4, 'omitnan'), class(FirstImage) );

If you do not have R2016a or later but you do have the Statistics Toolbox you can use

registered_image = cast( nanmean(temp_image, 4), class(FirstImage) );

Answer 2

[Adam Rauff]

This code generalizes this procedure to more than 2x2 (grayscale images only)

% define Y by X number of images
IMinfo.YXgrid = [10 8];

% chan4.IM is structure where images are stored in order
% i.e chan4(1).IM is image at the (1,1) position
%     chan4(2).IM is image at the (1,2) position
%     chan4(5).IM is image at the (2,1) position

% obtain final size of image
overlap = (Insert your overlap percentage);

[r, c] = size(firstImage); 
overlap_c = round(overlap * c); % pixels
new_c = c*IMinfo.XYGrid(1) - (IMinfo.XYGrid(1)-1)*overlap_c; % total columns in final image
overlap_r = round(overlap * r); % pixels
new_r = r*IMinfo.XYGrid(2) - (IMinfo.XYGrid(2)-1)*overlap_r; % total rows in final image

% intialize mosaic template
Mosaic = zeros(new_r, new_c);

% "stitch" images by overlaying them
for i = 1:IMinfo.XYGrid(2)
    roBegin = (i-1)*r + 1 - overlap_r*(i-1);
    roEnd = (i-1)*r + r - overlap_r*(i-1);
    for j = 1:IMinfo.XYGrid(1)
        colBegin = (j-1)*c + 1 - overlap_c*(j-1); % in matlab index begins at 1
        colEnd = (j-1)*c + c - overlap_c*(j-1);
        Mosaic(roBegin:roEnd, colBegin:colEnd) = chan4(j+((i-1)*IMinfo.XYGrid(1))).IM;
    end
end

[Timothy Sawe]

Hi Adam so I tried inserting my images using chan4 as below:

chan4(1).IM = imread('img1.jpg');
chan4(2).IM = imread('img2.jpg');

but it gives me an error:

Unable to perform assignment because the size of the left side is 1-by-6 and the size of the

right side is 2873-by-2825.

Error in pos_based (line 44)

Mosaic(roBegin:roEnd, colBegin:colEnd) = chan4(j+((i-1)*IMinfo.YXGrid(1))).IM;

I suspect I haven't inserted the images properly is why. How did you do it? P.S. They are in grayscale.

Answer 3

[Don Zheng]

Declare an image with the final size and register each of the four images according to your layout to the final image.