Matrix/Image Merging

Hi,

I have 2 matrix A and B which are different size, I would like to combine them together with A as the master to make a new matrix with consist of A and B. The way I want both matrix is merge through a location on both matrix. For example place Matrix A on Matrix B on certain location, Eg ‘8’ on Matrix A to Matrix B’s ‘22’ to result Matrix Merge.

A=[1 2 3 4 5; 6 7 8 9 10; 11 12 13 14 15]

A =

     1     2     3     4     5
     6     7     8     9    10
    11    12    13    14    15

B=[13 14 15 20 27 21; 16 17 18 19 25 26; 21 22 23 24 25 23]

B =

    13    14    15    20    27    21
    16    17    18    19    25    26
    21    22    23    24    25    23

Merge=[0 13 14 15 20 27 21 ; 1 2 3 4 5 25 26 ; 6 7 8 9 10 25 23 ; 11 12 13 14 15 0 0]

Merge =

     0    13    14    15    20    27    21
     1     2     3     4     5    25    26
     6     7     8     9    10    25    23
    11    12    13    14    15     0     0

Besides that, additional space will have value of 0. I'll use these method to merge 2 images where i have a location at both images which going to be used to control the location of merging.

Thanks;

[Kyle]

Problem solved by Sean de and Matt Fig.

Guys is it possible to modify the code to support this kind of matrix?

A=reshape(1:45,3,5,3);

B=reshape(1:105,5,7,3)+45;

The merging still same as before. Superimpose base location on Matrix A and Matrix B. However now there is 3 level of array. (i thinks its call 2 dimensional array, not very sure though)

i could store each level of array into separate 1 dimensional array and use the code u guys written to combine the matrix then put the matrix back into a new 2 dimensional array. But that means i need to run through the code 3 times.

Answer 1

[Sean de Wolski]

Repaste New trix every time!

clear Merge

A=[1 2 3 4 5; 6 7 8 9 10; 11 12 13 14 15];
B=[16 17 18 19 20 21; 22 23 24 25 26 27; 28 29 30 31 32 33; 34 35 36 37 38 39;40 41 42 43 44 45];

Awant = 8;
Bwant = 42; 

[ra ca] = find(A==Awant,1,'first');
[rb cb] = find(B==Bwant,1,'first');
D = abs(diff([ra ca;rb cb],1,1));
sd = sign(diff([ra ca;rb cb],1,1));
corners = abs(sum(sd));
if ~corners
if sd(1) == 1
    Merge(1:(size(B,1)),(D(2)+1:(size(B,2)+D(2)))) = B;
    Merge((D(1)+1):(size(A,1)+D(1)),1:(size(A,2))) = A;
else
    Merge((D(1)+1):(size(B,1)+D(1)),1:(size(B,2))) = B;
    Merge(1:(size(A,1)),(D(2)+1:(size(A,2)+D(2)))) = A;
end
elseif corners == 1;
if ~sd(1)
    Merge(1:(size(B,1)),1:(size(B,2))) = B;
    Merge((D(1)+1):(size(A,1)+D(1)),(D(2)+1:(size(A,2)+D(2)))) = A;
else
    Merge(1:(size(B,1)),1:(size(B,2))) = B;
    Merge((D(1)+1):(size(A,1)+D(1)),(D(2)+1:(size(A,2)+D(2)))) = A;
end    
else    
if sd(1) == 1
    Merge(1:(size(B,1)),1:(size(B,2))) = B;
    Merge((D(1)+1):(size(A,1)+D(1)),(D(2)+1:(size(A,2)+D(2)))) = A;
else
    Merge((D(1)+1):(size(B,1)+D(1)),(D(2)+1:(size(B,2)+D(2)))) = B;
    Merge(1:(size(A,1)),1:(size(A,2))) = A;
end
end

Should work for any case.

[Kyle]

When i try

Awant = 7;

Bwant = 22;

The result:

Merge =

13 14 15 20 27 21 0

16 1 2 3 4 5 0

21 6 7 8 9 10 0

0 11 12 13 14 15 0

But should be

C =

13 14 15 20 27 21

1 2 3 4 5 26

6 7 8 9 10 23

11 12 13 14 15 0

[Kyle]

C=[13 14 15 20 27 21; 1 2 3 4 5 26; 6 7 8 9 10 23; 11 12 13 14 15 0]

[Kyle]

Yep i tried clear it n it still give me

Merge =

13 14 15 20 27 21 0

16 1 2 3 4 5 0

21 6 7 8 9 10 0

0 11 12 13 14 15 0

[Kyle]

Weird then only thing i change from ur code is

Awant = 7;

Bwant = 22;

but still cant get the correct result like yours.

If u don't mind. Please re-paste ur code again

Thanks

[Kyle]

Thanks yep it work after u re paste.

I tested all possible configuration tat i could think of with ur code. The result is correct for all configuration except for one type.

When:

A=[1 2 3 4 5; 6 7 8 9 10; 11 12 13 14 15]

B=[16 17 18 19 20 21; 22 23 24 25 26 27; 28 29 30 31 32 33; 34 35 36 37 38 39;40 41 42 43 44 45]

Awant = 8;

Bwant =25; % try also 26,27

[Kyle]

Thanks a lot for ur help.

Hopefully u dont get offended. Found some error

Awant = 8;

Bwant =42; % 42,36,30, weird 23 also wrong

[Sean de Wolski]

Again!

You're doing your job of testing quite well. Better than me.

[Kyle]

Thanks.

No error found :D

Answer 2

[Matt Fig]

Kyle, what to do when multiple matches are found in B, since your example B has duplicates?

%

%

%

%

%

EDIT In response to clarification about duplicates.

Since you say there will not be duplicates in the actual data, I will use example matrices without duplicates:

A = reshape(1:12,3,4);
B = reshape(1:30,5,6)+12;

NA = 5;  % The number to overlap in A.
NB =  35;   % The number to overlap in B.

[mA,nA] = size(A);
[mB,nB] = size(B);
[IA,JA] = find(A==NA);
[IB,JB] = find(B==NB);
mC = mA+mB+mod(mA+mB,2)+1;
nC = nA+nB+mod(nA+nB,2)+1;
C = zeros(mC,nC);
cC = round([mC/2,nC/2]);
C(cC(1)-IB+1:cC(1)-IB+mB,cC(2)-JB+1:cC(2)-JB+nB) = B;
C(cC(1)-IA+1:cC(1)-IA+mA,cC(2)-JA+1:cC(2)-JA+nA) = A;
C = C(:,any(logical(C)));
C = C(any(logical(C),2),:)

[Kyle]

For Sean's code, he find the location of the Bwant then start computation to determine how to superimpose A on to B.

Therefore it doesnt matter if it has multiple matches as long as u know the location of the point in A and B

[Matt Fig]

But if there are two matches, then there is two ways to overlap. Which is preferred?

[Kyle]

Oh, i'm going to use this concept to merge 2 image. So there would only be 1 match on each image.

I tried this concept on gray scale image and its applicable. Now i'm trying to used it on RGB image as RGB has 3 array in 1 single image.

A(:,:,1:3)

[Sean de Wolski]

That would crop out black columns or rows of an image.

[Matt Fig]

True, but Kyle said there were no duplicates. In that case there shouldn't be an entire row or column that is the same!

[Sean de Wolski]

Good Point.

[Kyle]

Ur code does work for matrix. Very robust

But when i applied on gray images. The output image doesnt seems to be from my input image

Answer 3

[Kyle]

Hi Matt,

i used ur code to test on image like shown below.

clc

% A = reshape(1:15,3,5)
% B = reshape(1:35,5,7)+12

A = imread('cameraman.tif');
B = imread('cameraman.tif');

% NA = 8;  % The number to overlap in A.
% NB =  32;   % The number to overlap in B.

[mA,nA] = size(A);
[mB,nB] = size(B);

% [IA,JA] = find(A==NA);
% [IB,JB] = find(B==NB);

IA=50;
JA=50;

IB=1;
JB=1;

mC = mA+mB+mod(mA+mB,2)+1;
nC = nA+nB+mod(nA+nB,2)+1;
C = zeros(mC,nC);
cC = round([mC/2,nC/2]);
C(cC(1)-IB+1:cC(1)-IB+mB,cC(2)-JB+1:cC(2)-JB+nB) = B;
C(cC(1)-IA+1:cC(1)-IA+mA,cC(2)-JA+1:cC(2)-JA+nA) = A;
C = C(:,any(logical(C)));
C = C(any(logical(C),2),:);

imshow(C)

The resultant image should be Image A overlapping Image B. however i only see black n white. Did i do anything wrong? Ur code works on matrix, and image is also a form of matrix. I dont know how come it when wrong.

[Matt Fig]

I don't have IMSHOW (I think that is an image processing toolbox function), but it works with:

image(C)

colormap(gray)

You will have to figure out what went wrong with IMSHOW, as I have no way to trouble-shoot.

[Kyle]

This is odd. i even check the pixel value. its the same but shows different color. First time encounter this

[Kyle]

http://www.mathworks.com/matlabcentral/answers/10268-weird-imshow-image-same-pixel-value-different-color

Problem solved, need to change this

C = zeros(mC,nC,'uint8');