How can i implement this code?

Hi

I'm looking for a matlab code which estimates the two principal directions (axes) using principal components analysis (PCA)?? Then any an other algorithm can be applied to estimate the asymmetry of the region of image in terms of the two principal axis. I tried several times to implement it and I looked everywhere to no avail. Now I ask my question on this forum hoping to get the right answer. thanks

Answer 1

[Sachin Ganjare]

Refer below link:

http://matlabdatamining.blogspot.jp/2010/02/principal-components-analysis.html

Hope it will be useful.

[Tomas]

thank you for this link. I visited it but since I'm a beginner in matlab I have not got to use these concepts to draw the axes of the region in the segmented image and estimate the asymmetry.

Answer 2

[Tomas]

I tried to do this code but i need your help please

I2=segmentation(I);
[n m] = size(I2);
AMean = mean(I2);
co=I2*I2'; % Compute the covariance matrix (co)
% Compute the eigen values and eigen vectors of the covariance matrix
[eigvector,eigvl]=eig(co);
eigvalue = diag(eigvl);
pc = eigvector * I2;
plot(pc(1,:),pc(2,:))

[Matt Kindig]

What is your question? What about this code doesn't work?

[Tomas]

Hi

I'm looking for a matlab code which estimates the two principal directions (axes) using principal components analysis (PCA)?? Then any an other algorithm can be applied to estimate the asymmetry of the region of interest of the segmented image in terms of the two principal axis. I tried to do this code but it doesn't work so i need your help.

Answer 3

[Tomas]

Hi, Can any one please correct this code

A=imread('aeroplane silhouette.png');
bw=~im2bw(A,0.5); %Threshold and invert
subplot(1,2,1), imshow(bw,[]);
[y,x]=find(bw>0.5); %Get coordinates of non zero pixels
centroid=mean([x y]); %Get (centroid) of data
hold on; plot(centroid(1),centroid(2),'rd'); %Plot shape centroid
C=cov([x y]); %Calculate covariance of coordinates
[U,S]=eig(C)
m=U(2,1)./U(1,1);
const=centroid(2)/m.*centroid(1);
xl=50:450; yl=m.*xl+const
subplot(1,2,2), imshow(bw,[]); h=line(xl,yl); %Display image and axes
set(h,'Color',[1 0 0],'LineWidth',2.0)
m2=U(2,2)./U(1,2);
const=centroid(2)/m2.*centroid(1);
x2=50:450; y2=m2.*x2+const
h=line(x2,y2);
set(h,'Color',[1 0 0], 'LineWidth',2.0)

you find the images here

http://img15.hostingpics.net/pics/679738aeroplanesilhouette.png

http://img15.hostingpics.net/pics/171229plane.png

[Matt Kindig]

Before asking us to correct the code, you should tell us what's wrong with the code. From the second image, it seems that the gray lines lie along what I would expect to be the principal components.

[Tomas]

These image explained what i should to do but this code didn't plot the two axes