The following are the questions:
- Implement down-sampling with anti-aliasing using the Gaussian LPF (hint: reduce the house grayscale image four times, first by applying a Gaussian filter and then by filtering every other row and column. Compare the output images with and without pre-processing with LPF before down-sampling).
- Use the FFT to up-sample an image: first double the size of the
lenagrayscale image by padding zero rows/columns at every alternate positions, then use the FFT followed by an LPF and then by the IFFT to get the output image. Why does it work? - Try to apply the Fourier transform and image reconstruction with a color (RGB) image. (Hint: apply the FFT for each channel separately).
- Show (mathematically and with a 2D kernel example) that the Fourier transform of a Gaussian kernel is another Gaussian kernel.
- Use the
lenaimage and the asymmetric ripple kernel to generate images with correlation and convolution. Show that output images are different. Now, flip the kernel...
