First of all, let me say Go Coogs.
Second, here's some code you can use. This works from the input image provided, i.e., it first extracts the rectangular black regions containing the letters U and H, then combines them into a new image and colors the letters red.
(If instead you started from two input images just containing the rectangular regions containing the letters, then the first part of this where the U and H are extracted could be skipped.)
im_input = imread('input.png');
[m,n,p] = size(im_input);
% first, have to extract the rectangles (almost squares - so I'll call
% them squares) containing the letters from the input image:
% find the upper-left corner of the U square by finding the first black
% pixel in the image:
[u_start_row,u_start_col] = find(all(im_input == 0,3),1)
% find the lower-right corner of the U square by finding the last black
% pixel in the left half of the image:
[u_end_row,u_end_col] = find(all(im_input(:,1:floor(n/2),:) == 0,3),1,'last')
% find the upper-left corner of the H square by finding the first black
% pixel in the right half of the image, then adjust the column index
% accordingly:
[h_start_row,h_start_col] = find(all(im_input(:,floor(n/2)+1:end,:) == 0,3),1);
h_start_row
h_start_col = h_start_col+floor(n/2)
% find the lower-right corner of the H square by finding the last black
% pixel in the image:
[h_end_row,h_end_col] = find(all(im_input == 0,3),1,'last')
% copy the black squares containing the letters to separate variables:
im_u = im_input(u_start_row:u_end_row,u_start_col:u_end_col,:);
im_h = im_input(h_start_row:h_end_row,h_start_col:h_end_col,:);
% show them:
imshow(im_u);
imshow(im_h);
% make sure the "squares" are the same size
% (if not, then one or both needs to be adjusted by adding black rows and/or
% columns to match the other one's size, but here they are the same size,
% luckily.)
assert(isequal(size(im_u),size(im_h)))
% now put them together and change the letters into red:
% first, have to decide how to consider whether a pixel is part of a
% letter (2 options):
% option 1: to use the locations of all pixels that aren't black in the
% final output (i.e., include the grey borders around the letters):
% u_idx = ~all(im_u == 0,3);
% h_idx = ~all(im_h == 0,3);
% option 2: to use the locations of the white pixels only in the final
% output image (i.e., excluding the grey borders around the letters):
% I think this is how the actual output image is.
u_idx = all(im_u == 255,3);
h_idx = all(im_h == 255,3);
% finally, construct the output image. first as a cell array containing a
% 2D matrix for each channel (R,G,B):
im_uh = num2cell(zeros(size(im_u),'uint8'),[1 2]);
% red color:
c = [255 0 0];
% for each channel ...
for ii = 1:p
% ... make the pixels red where they're part of a letter:
im_uh{ii}(u_idx | h_idx) = c(ii);
end
% combine the (R,G,B) cell array into a 3D matrix:
im_uh = cat(3,im_uh{:});
% show the result:
imshow(im_uh); % Go Coogs!
