ECE 563 Assignment 3

Images and m-files for this assignment may be downloaded from impro3.zip and head.zip.

Individual Work

  1. Select a color photograph, with lots of bright colors. It should be a photograph, not a magazine or book image. Scan the photograph twice at the same resolution (70-150 dpi), once at 24-bit color and once at 8-bit grayscale. Be careful not to move the photograph between scans. Load both images into Matlab. Convert the color image to grayscale (rgb2gray) and generate a histogram of the difference between that image and the scanned grayscale image. Display a scaled 'difference' image obtained by mapping the image such that positive values range from (0.5 to 1) or negative values range from (0 to 0.5) depending on which range is larger. See kitchen1.

  2. Using your scanned photograph, determine the coefficients in the equation Y = a0 + a1*X, where X is the scanned grayscale image and Y is the color image converted to grayscale by rgb2gray. Calculate the correlation between the two images before and after linear correction.

  3. Using the same photograph, determine the coefficients in the equation Y = k1 R + k2 G + k3 B by the method of least squares where (R, G, B) are values from the 24-bit digitized image and Y are values from the 8-bit grayscale digitized image. Subtract the image resulting from your fit from the grayscale image and generate a histogram of the differences. Find the maximum deviations and the rms deviation of the error. Scale the difference image to full grayscale range and examine the result. Does the difference image correlate with the the original image?

  4. Photograph (or scan) a collection of objects that would be suitable for binary image analysis of shapes. Suggestions include keys, coins, washers, stamps, leaves, and puzzle pieces. Show the histogram of your image and select an appropriate threshold using (a) Matlab threshold function and (b) manual selection.

Group Projects

  1. Do textbook problems 10.20, 10.22, 10.23, 10.24, and 10.25

  2. Generate 10 uniform random images. Average the first n images, where n = 1 .. 10. Display the results (image and histogram) and calculate the mean and standard deviation of the resulting images. Explain what is happening.

  3. The image graycard_75dpi.tif is a scanned image of a Kodak gray reference card. Find the mean and standard deviation of the image. Generate a gaussian random image with the same mean and standard deviation. Display the two images side by side and assess their similiarity or difference.

  4. Generate a binary image with an array of regular polygons with 3 to 8 sides. You may use polygon.m for this purpose.

  5. For the binary images of a spade, club, heart, and diamond, generate a table of features, to include the centroid, rectangular extent (width and height), area, perimeter, and circularity. You may find the Matlab routine regionprops useful.

    Reference:

    William Pratt, Digital Image Processing, (Third Ed)
    Wiley-Interscience, 2001. p 596-597.

  6. Generate a cross-section of the MRI head as shown below

    Generate a movie of the MRI head starting from image 127 downto image 0.

Maintained by John Loomis, last updated 22 Jan 2005