ECE 563 Assignment 14 (due 21 April 2004)
Individual Projects
- Using a 24-bit color image, jpeg compress the image using quality factors
of (5%, 10%, 15%, 25%, 50%, 75%, and 90%). Tabulate the file sizes and compression ratio
achieved. Show the lower-quality images on your web page and compare to the
original. Then convert the original image to grayscale repeat the compression
versus jpeg quality factor table.
Are the grayscale images 1/3 the size of the color images?
- Using the multi-level quad transform method shown in class
(download splitter.zip), transform a
sample color image and show that you can recover the original image.
Then take the transformed image and convert it to jpeg for n =
1:8 levels (eight different results). Tabulate (and plot) the size of
the resulting jpeg files vs the number of levels. Read each jpeg
image back and try to recover the original image. Tabulate (and
plot) the maximum and rms differences between the restored and
original images vs. the number of levels. The original image should
be a power of two and larger than 256 x 256.
Group Projects
- Assemble the images in seq.zip into a single
mosaic by developing an automated matlab function to determine the overlap
offset between images.
Here is a manual attempt at the mosaic:
- When WinZip compressed the image files in impro14.zip, the following
compression ratios were achieved.
| name | Length | Size | ratio |
|---|
| svoids.tif | 124,764 | 82,976 | 1.504 |
| ape.tif | 15,678 | 7,816 | 2.006 |
| tbrawl.tif | 312,722 | 63,309 | 4.940 |
| lighthou.tif | 78,854 | 63,615 | 1.240 |
| bears.tif | 312,722 | 208,328 | 1.501 |
| xray.tif | 78,854 | 54,756 | 1.440 |
| truck.tif | 695,940 | 530,742 | 1.311 |
Modify this table to show the image size (rows x cols) and pixel depth (bits-per-pixel) for
each image.
- Convert the files above into gif and jpeg formats. Use 75% quality for the
jpeg images (the default). Add the file sizes and compression ratios to the table above (or start
a new table). Which compresses better - gif, jpeg or zip?
- Using a grayscale image (such as lighthou.tif), jpeg compress the image using quality factors
of (5%, 10%, 15%, 75%, and 90%). Tabulate the file sizes, compression ratio, and deviations (max,
min, and rms) between the compressed image and the original image. Prepare an animated sequence
showing the degradation of the jpeg as a function of quality factor.
- Obtain the Haar transformation matrix for N = 8 and do the following:
- Compute the Haar transform T = H F H of the 8 x 8 matrix F = magic(8).
- Compute the inverse Haar transform F = H-1 T H-1.
- Verify that the wavelet transform and inverse transform (download wavelet.zip)
restores the original image
for db4 (group1), sym4 (group2), bior6.8 (group3), and jpeg9.7 (group4) for
n = 1:4 levels.
Maintained by John Loomis,
last updated 10 April 2004