Here I blurred an image by taking the FFT of a filter and of the image itself and multiplying them together, then I took the inverse transform to get the original image.
The following is an image of the fft of the gaussian used for the
following images, with the fftshift function applied:
This method simply convolves the image with a Gaussian filter directly, which produces similar results to convolving the FFT of the image with the FFT of a Gaussian filter. Note: I did not follow the hint in order to more accurately show the equivalence of the two techniques. The reason they both have the same effect is that multiplication in the Fourier space is equivalent to convolution in the image space.
These images were sharpened using
the fspecial('unsharp') filter generated by matlab. I
found that in many cases this had the effect of amplifying noise in
the image, but it was able to emphasize some of the details.
I generated a "ghost" effect in an image by overlaying a partially transparent black and white version of the "ghost" with the original background image. I also blurred the images a little to make it look more "ghosty".
Here we brought out the darker parts of the image
using histeq, which helps us to see the differences in
the lighter colors more easily. This works on both black and white and
color images. This worked well to reveal information on an image of a
nightclub that was mostly dark. It didn't work quite as well on an
image of Pittsburgh which also contained a lot of bright parts.
