Image Processing - MATLAB Recipes for Earth Sciences - page 326

median-i ltered version of the I2 image is, of course, very smooth compared
to the original I1 image. We would, however, lose a lot of detail if we used
this version of the image. We next subtract the median-i ltered image I2
from original image I1 , which yields the image I3 .
I3 = imsubtract(I1,I2);
imshow(I3,'InitialMagnification',200)
title('I1-I2')
We then subtract the original image I1 from the median-i ltered image I2 ,
which yields the image I4 .
I4 = imsubtract(I2,I1);
imshow(I4,'InitialMagnification',200)
title('I2-I1')
We next replace the original pixels with their median-i ltered versions if the
dif erence between the median-i ltered image I2 and the original image I1 is
great than 10 in both directions (as it is in our example).
I5 = I1;
I5(I3>10 | I4>10) = I2(I3>10 | I4>10);
imshow(I5,'InitialMagnification',200)
title('Despeckled Image')
h e image I5 obtained using this approach is the despeckled version of the
image I1 . We can also explore the pixel values of both versions of the image
( I1 and I5 ) in a 3D surface plot, using
subplot(1,2,1)
I1S = im2double(I1);
surface(I1S), colormap jet, caxis([0 1])
shading interp, view(120,33), axis off
axis([1 size(I1,1) 1 size(I1,2) min(I1S(:)) max(I1S(:))])
subplot(1,2,2)
I5S = im2double(I5);
surface(I5S), colormap jet, caxis([0 1])
shading interp, view(120,33), axis off
axis([1 size(I1,1) 1 size(I1,2) min(I1S(:)) max(I1S(:))])
We need convert the image data to class double using im2double in order to be
able to display the data using surface . Finally, we can display both images in
the same i gure window
subplot(1,2,1), imshow(I1), title('Original Image')
subplot(1,2,2), imshow(I5), title('Despeckled Image')
to see the result of despeckling the image I1 (Fig. 8.2).