How to take the mean of hyperspectral image of size (1312*959*31) and plot its spectrum in matlab 7.0 after masking the specific part ?

S=load('Daylight_Scene_06.mat');
image1=S.tensor;
image1=S.tensor(:,:,1);  %S.tensor(:,:,k) k wavelengths
imshow(image1);
%BW=roipoly;
BW = roipoly(image1);
r=zeros(1312,959);

for i=1:1312;
    for j=1:959;
        if BW(i,j)==1
            r(i,j)=image1(i,j);
             %r2(i,j)=image2(i,j);
            end
     end
end
imshow(r);

image2=S.tensor;
image2=S.tensor(:,:,2);  %S.tensor(:,:,k) k wavelengths
imshow(image2);
%BW=roipoly;
BW = roipoly(image2);
r1=zeros(1312,959);
for i=1:1312;
    for j=1:959;
        if BW(i,j)==1
            r1(i,j)=image2(i,j);
             %r2(i,j)=image2(i,j);
           end
     end
end
imshow(r1);

Answer 1

[Guillaume]

Learn to use matrix operations instead of loops. Your code will be a lot faster and much easier to read. For example, your first loop and the declaration of r can be replaced by just this one line:

r  = image2 .* (BW == 1); 
%and if BW is just 0 and 1, then:
% r = image2 .* BW;

The same can be applied all at once to your hyperspectral image with:

maskedimage  = bsxfun(@times, S.tensor, BW == 1); 
%and if BW is just 0 and 1, then:
%maskedimage  = bsxfun(@times, S.tensor, BW);

To average all the images into one:

meanmaskedimage = mean(maskedimage, 3); %average hyperspectral image along the 3rd dimension

[Rashmi]

Thanks for the answer. I want the mean of roi of every band.please elaborate ans.

[Guillaume]

In your example, it appears that you acquire a different roi by hand for the first two bands. Are you planning to do that for each band?

Assuming not, and that you have acquired a single roi any way you want, then the masking is exactly what I showed:

%BW is a roi image, acquired any way you want, e.g.
%by calling roipoly on the first band:
%BW = roipoly(S.tensor(:, :, 1));

%Apply roi masking to all bands:
maskedbands  = bsxfun(@times, S.tensor, BW);

To calculate the mean of each band, you can either first calculate the mean of the rows, then the column mean of that:

meanbands = squeeze(mean(mean(maskedbands), 2));

Or you can reshape the 3D matrix into 2D where each column consists of all the pixels of each band:

meanband = mean(reshape(maskedbands, [], size(maskedbands, 3))