ECE 563 Assignment 13 (due 14 April 2004)
Images for this assignment may be downloaded from
impro13.zip.
Individual Projects
- Find the autocorrelation and selfconvolution images of an unsymmetric shape
of your choice.
(See example)
- Generate a uniform random (0-1) image the same size as your class picture (convert to grayscale).
Smooth the random image with a gaussian filter whose width is 10-25% of the image. Multiply the Fourier Transform
of your class image by the resulting random image applied as a phase shift.
(i.e. exp(-2 j p k R(x,y)), where k is a scale factor,
and R(x,y) is the random image.) Inverse transform the results and show the magnitude of
the resulting images for different scale factors.
- Use grayscale images of two class members as input. Generate their Fourier transforms.
Interchange the amplitude and phase terms of the two images and generate the corresponding
inverse Fourier transform images.
Group Projects
- Crop out a rectangular sample from
xruler.tif,
showing a millimeter scale and such that the image dimensions are
powers of two (e.g. 256 x 16). Determine the ruler frequency from
the Fourier transform of this sample (measure the distance from the
central peak (DC term) to the first strong harmonic peak). Compare
the results to those determined by measuring the period (scale) directly.
- Rotate a small image using imrotate. Verify that the
Fourier transform image rotates by the same amount by showing
the rotated image and its Fourier Transform. Demonstrate with an animation.
- For the images of letters of the alphabet, shown below, prepare an array of images showing the
cross correlation and cross convolution images of each letter with itself and the other two letters.
The diagonal of this array will be autocorrelations or self convolutions.
 
- Filter the spurious patterns in the lincoln image.
- Use an angle-selective low-pass filter on the fabric image below
to remove the structure of
the individual strands and fibers. Note that most of the strands have
frequencies in the horizontal and vertical directions, whereas the weave is
oriented along the 45-degree axes.

- Prepare powerpoint tutorials on the following Matlab topics:
- group 1 - Radon transform
- group 2 - Fan-beam projection
- group 3 - Discrete Cosine Transform (DCT)
- group 4 - DCT in JPEG image compression
Maintained by John Loomis,
last updated 7 April 2004