Lets Learn together... Happy Reading

" Two roads diverged in a wood, and I,
I took the one less traveled by,
And that has made all the difference "-Robert Frost

Extraction of Connected components without using BWLABEL function


             Many of the visitors to this blog mailed me to post a MATLAB code for extracting the connected components.  In MATLAB, a function called BWLABEL is available to label the connected components.  
Based on the following iterative expression, the connected components are extracted.
where k=1,2,3…

Using the above expression, connected components are extracted without using the function BWLABEL
LABELLING:

MATLAB CODE:


I=imread('label3.jpg');





I=im2bw(I);

%Structuring element
B=strel('square',3);
A=I;
%Find a non-zero element's position.
p=find(A==1);
p=p(1);
Label=zeros([size(A,1) size(A,2)]);
N=0;

while(~isempty(p))
    N=N+1;%Label for each component
    p=p(1);
X=false([size(A,1) size(A,2)]);
X(p)=1;

Y=A&imdilate(X,B);
while(~isequal(X,Y))
    X=Y;
    Y=A&imdilate(X,B);
end

Pos=find(Y==1);

A(Pos)=0;
%Label the components
Label(Pos)=N;

p=find(A==1);

end
imtool(Label);












EXPLANATION:
  1. Read an image (A) and convert it into binary image.
  2. Define a structuring element (B).
  3. Initialize the Label matrix with zeros.
  4. Find a non-zero element position in the input matrix A.
  5. Initialize a matrix X with zeros and place 1 in the non-zero element position found in the previous step.
  6. Perform dilation using the structuring element B on matrix X. i.e. imdilate(X,B);
  7. Perform intersection with the matrix A. Y= A&imdilate(X, B).
  8. Check whether Y==X. If no, then X=Y and perform steps 6 and 7 again else stop the iteration.
  9. Find the non-zero elements position in the Y. In matrix Label place a number N in those positions. N is for labeling the connected components.
  10. Similarly, place zero in those positions in the input matrix A.
  11. Again find a non-zero element position in the matrix A. If found, goto step 5 else stop the iteration.
  12. Using the labels the connected components can be extracted.


EXTRACTION:

MATLAB CODE:






%Extracting the components
Im=zeros([size(A,1) size(A,2)]);
ele=find(Label==1);
Im(ele)=1;
figure,imshow(Im);title('Label:1');






To obtain the first component, find the positions with value=1 in the Label Matrix. Similarly, other components can be extracted.




%Extracting the characters 'I M A G E'
ele=find(Label==2|Label==3|Label==6|Label==7|Label==9);
Im1=zeros([size(A,1) size(A,2)]);
Im1(ele)=1;
figure,imshow(Im1);title('Specific components');






From the Label matrix, I extracted the characters I, M, A, G and E alone by specifying the label numbers 2,3,6,7 and 9.


%Total number of Letters



Total=sprintf('Total Number of Letters:%d',N);
display(Total);




 The total number of components extracted in the above example Image:

Total =

Total Number of letters:14




%Differentiate each component with a specific color
RGBIm=zeros([size(Label,1) size(Label,2) 3]);
R=zeros([size(Label,1) size(Label,2)]);
G=zeros([size(Label,1) size(Label,2)]);
B=zeros([size(Label,1) size(Label,2)]);
U=64;
V=255;
W=128;
for i=1:N
    Pos=find(Label==i);
    R(Pos)=mod(i,2)*V;
    G(Pos)=mod(i,5)*U;
    B(Pos)=mod(i,3)*W;
   
   
 end
RGBIm(:,:,1)=R;
RGBIm(:,:,2)=G;
RGBIm(:,:,3)=B;
RGBIm=uint8(RGBIm);
figure,imshow(RGBIm);title('Labelled Components');






For each label, unique color is given.  Here is another example for labeling and extracting the components. Read the image, ‘coins.png’ , convert it into binary and fill the holes. Then perform labeling and extraction.





Total =

Total Number of coins: 10



like button Like "IMAGE PROCESSING" page

7 comments:

Unknown said... Reply to comment

please explain th coloring section of the code.
If i need to color all the labelled objects as only "black" or "white" , depending on the background how can i do it..?

Aaron Angel said... Reply to comment

@Pratik Datta

Coloring is done based on the index of the objects. Each component(pixels with same index are assumed as one component) is assigned one color.
For more information:
'label2rgb' is a matlab function that performs similar task.

Unknown said... Reply to comment

Thank you for this code, you are a life saver

Unknown said... Reply to comment

how can i extract without knowing label number?

rs said... Reply to comment

Is it possible to implement this without matlab, in C++?

Unknown said... Reply to comment

Sir,how text localization can be done

Maryam Naseer said... Reply to comment

Great work!

Enjoyed Reading? Share Your Views

Previous Post Next Post Home
Google ping Hypersmash.com