ECE 563 Assignment 8 (due 3 March 2004)

Images and matlab functions for this assignment may be downloaded from impro8.zip.

Individual Work

  1. Choose an image, convert it to grayscale, and display (a) the original grayscale image (b) magnitude of gradient image, and (c) Laplacian filtered image.

Group Projects

  1. Repeat the steps in the individual problem for the fruit images from assignment 6 and the euclidian and Mahalanobis distance images (apples and oranges).

  2. Take the Washington DC sequence from assignment 5. Histogram equalize each band separately. Combine the first three components into an RGB image. Compare to similar images from assignments 5 and 6. Find the principal components of the histogram-equalized bands. How have the eigenvalues and eigenvectors been affected?

  3. Generate progressively blurred sets of images, similar to that shown below, using different averaging filters (e.g. flat, gaussian, binomial). To make the problem current, create an image of George Bush's head that is about the same size as Clinton's head image. The handout has a possible starting image (bush2.tif).

  4. Let a gaussian image with standard deviation s1 be filtered with a gaussian averaging filter with standard deviation s2. Find the standard deviation of the resulting image. Do analyically and verify numerically.

  5. Calculate the mean and standard deviation of graycard_75dpi (from impro3.zip) as a function of the number of times a certain averaging filter is applied (flat, gaussian, binomial).

  6. Generate a set of grayscale images using increasing amounts of salt-and-pepper noise (using the matlab function imnoise). Show the improvement (if any) after filtering with smoothing filters and median filters (see medfilt2). Evaluate the degree of improvment obtained, by calculating the statistics of the differences between the original image and both the degraded and improved images.

  7. Find third and fourth derivative filters by repeated convolution of [-1 1]. Find an orthonormal set of 1 x 5 filters such that each filter represents a successive derivative. Show the results of applying these filters on the set of coordinate images, [ 1 x x2 x3 x4].

Maintained by John Loomis, last updated 24 Feb 2004