How to convert 3D matrix to 1D?

147 visualizzazioni (ultimi 30 giorni)
Tan Wen Kun
Tan Wen Kun il 3 Dic 2015
Modificato: Guillaume il 4 Dic 2015
I got a color image and store into a variable A.
A =
(22,10,10) (22,10,10) (39,40,50)
(89,11,14) (23,11,11) (99,10,10)
(89,11,14) (69,10,10) (99,10,10)
x=1 when loop horizontally I wish to register the first color I seem and label it into x.
when loop same color then label it into same label.
when new color is seem then register the color and label into x+1.
I wish to get a table like
color label
(22,10,10) 1
(39,40,50) 2
(89,11,14) 3
(23,11,11) 4
(99,10,10) 5
(69,10,10) 6
so result finally get is
B=
1 1 2
3 4 5
3 6 5
How to get the label table and how to get the result?
  2 Commenti
Walter Roberson
Walter Roberson il 3 Dic 2015
Modificato: Walter Roberson il 3 Dic 2015
What data type is A? The notation you use is not MATLAB syntax, and
A = {
[22,10,10], [22,10,10], [39,40,50];
[89,11,14], [23,11,11], [99,10,10];
[89,11,14], [69,10,10], [99,10,10] }
would not be a 3D matrix.
Tan Wen Kun
Tan Wen Kun il 3 Dic 2015
Modificato: Tan Wen Kun il 3 Dic 2015
I should say I got a 3D image.
I got a color image like this so what I want to do is loop horizontal and label the color seem and register it.
I just show my problem in a simple form to see.
so I did not use
A(:,:,1) =
22 22 39
89 23 99
89 69 99
A(:,:,2)=
10 10 40
11 11 10
11 10 10
A(:,:,3)=
10 10 50
14 11 10
14 10 10

Accedi per commentare.

Risposta accettata

Guillaume
Guillaume il 3 Dic 2015
The way to do this is to reshape your image into an nx3 matrix where each row correspond to the three colours of a pixel. You can then apply unique with the 'rows' option on this to get your labels. It's then just a matter of reshaping the output.
Note that doing the scanning by rows rather than by columns complicate the code somewhat.
Also, please, post valid matlab code rather than leaving it to us to generate valid matrix
A = cat(3, [22 22 39;89 23 99; 89 69 99], ... it would have been great
[10 10 40; 11 11 10; 11 10 10], ... if I could just have
[10 10 50; 14 11 10; 14 10 10]) %copy/pasted that from your question
[~, ~, labels] = unique(reshape(permute(A, [2 1 3]), [], 3), 'rows', 'stable');
B = reshape(labels, size(A, 2), size(A, 1))'
The same code if you do the labeling by column:
[~, ~, labels] = unique(reshape(A, [], 3), 'rows', 'stable'); %no need to transpose anymore
B = reshape(labels, size(A, 1), size(A, 2)) %no need to transpose either.

Più risposte (1)

Walter Roberson
Walter Roberson il 3 Dic 2015
If you do not need the labels to start from 1 at the upper left, then
[color, ~, idx] = unique(reshape(A, [], 3), 'rows');
B = reshape(idx, size(A,1), size(A,2));
  9 Commenti
Tan Wen Kun
Tan Wen Kun il 3 Dic 2015
Is that other method can do like stable? I hope to get 1 at upper left
Columns 1 through 10
152 152 152 152 152 152 152 152 152 152
Columns 11 through 20
152 152 152 152 152 152 152 152 152 152
Columns 21 through 30
152 152 152 152 152 152 152 152 152 152
Guillaume
Guillaume il 3 Dic 2015
Without 'stable', it's not possible to have a row or column ordering of labels with unique. You would have to use sort with the stable option (I'm fairly certain that did exist in 2011b) and the find the duplicate yourself. It's possible but it's worthy of a question in itself.
As for your new question, please start a new question as it is barely related to the current one. When you ask the question, please explain clearly what your definition of connected is. In your above example, I consider 4 and 5 to be connected orthogonally, and 3 and 5 to be connected diagonally.

Accedi per commentare.

Tag

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by