ECE 563 Assignment 7 (due 25 Feb 2004)

Images and matlab functions for this assignment may be downloaded from impro7.zip.

Individual Work

  1. Histogram equalize your digitized color photograph in two different ways. First, histogram equalize the red, green, and blue images separately. Second, histogram equalize the grayscale version of your image, and apply the identical histogram modification to the red, green, and blue components of your image.

  2. The human face is (approximately) bilaterally symmetric. Use a frontal face image of yourself, separate the two halves, and generate left and right symmetric face images using each half and its reflection.

  3. Download a series of consecutive images from a web camera. Analyze the differences between several of the images to determine what features of the image are changing (traffic, pedestrians, boats, etc). Produce a movie or animated gif of the result. See the examples in Webcam sequence, Stadium and Cincinnati. Produce a baseline "empty" image by combining three or more images, using only those areas that are static (i.e., present in the majority of the images).

Group Projects

  1. Tabulate the image statistics using statxture (see p 464-467 and impro4.zip) for the noise images generated in assignment 4. Compare the results numerically to those found in imstats (in impro7.zip) and explain any differences in values. Prepare postage-stamp images (say 64 x 64) of the histograms of each noise image, and arrange the images in rows (or columns) of increasing value of various statistical parameters (skewness, smoothness, kurtosis, uniformity, entropy). You may find histogram_image.m from impro7.zip a useful model.

  2. Prepare a series of images to document the various pixel transformation methods in intrans.m and gscale.m from the textbook (see p 68-76 and 573-574).

  3. Use histogram equalization on tools.jpg, breast.tif (see p 68), bone-scan-GE.tif (see p 75), polen.tif (see p 83),chestXray.tif (see p 139), and on one of the MRI images of the head cross-section from assignment 3. Show before-and-after images and histograms.

  4. Calculate the invariant moments (p 470-474) of the playing card images from assignment 3. Use the matlab file invariant_moments.m in the handout. Compare the results to those generated by invmoments.m (p 576-578). Report and resolve any differences (errors) you find in those files.

  5. Calculate the invariant moments of binary_blobs.tif. Segment the image so that you process only one blob at a time. Are the moments invariant to rotation?

  6. Use Fourier descriptors (p 458-462) to generate images from chromo_binary.tif like those on page 462). The program frdescp (page 459-460) was scanned for assignment 6. Prepare similar sets of images for the playing card images.

  7. Prepare a powerpoint tutorial on the use of the watershed algorithm (image processing toolbox). We will be working on applications of this algorithm in the next assignment

Maintained by John Loomis, last updated 16 Feb 2004