(1) (2 points) Create an RGB image that is 2×2 where the upper left pixel is red, the upper right
pixel is green and lower pixels are blue. No need to plot the image, but print out the values in the
variable that makes up the image.
(2) For this problem I recommend first working through the Practical Lesson, 3.7.5. Instead of
using the chicken images, you will use homework_images.mat. When you read the file in, the
variable name for the array will bucky_image and this is a 3 dimensional array with 8 Bucky
images with varying amounts of noise along the 3rd dimension. Sort the 8 images according to
SNR and then CNR. I do not like the trick the author used to sort, since SNR is rounded and in
many images I’ve worked with the SNR can be very low and rounding would set all SNRs to 0. In
fact, in this case you will not be able to sort the images well if you round. Instead, use the sort
function in MATLAB. Also, cut down on the amount of looping used in the author’s example. You
(a) (2 points) Print the unordered SNRs (according to the order of the images).
(b) (1 point) Plot the Bucky images in a 2 × 4 grid of images, ordering the images from lowest
to highest SNR. Display in grayscale.
(c) (2 points) Repeat the above for CNR. Print out the CNRs and plot the images, ordered by
CNR. Did the order change? Explain why it did or did not change.
(3) For this problem use the image bucky_trio.mat that was distributed with this assignment.
Please use grayscale for all parts. This image contains 3 repetitions of our school mascot in a single
image, although all three may not be visible at once when viewing.
(a) (1 point) Display the image in grayscale, properly, using image.
(b) (4 points) Create a 3 × 1 panel plot where panels 1, 2 and 3 display processed image results,
using methods from Chapter 4, that reveal the first, second and third Bucky, respectively.
The full 351 × 1920 image should be plotted in each panel. I expect you to use a structured
procedure for pulling out each image, as opposed to guessing many numbers until one works.
You should be able to make an educated, very good, first try due to the regular structure of
the image portions you are trying to extract.
(c) (1 point) When you did the previous part each of the three portions of the image containing
the Buckys changed intensity. Describe how they change and why they changed in that way.
(4) For this problem you will use img_029.ppm from the distributed final project data (available
to download in the Canvas Files, distributed.zip). The goal of the final project is to pull out
features of images that help indicate when a retina shows retinopathy. This is a clear example of
such a retina, as indicated by the lighter colored spots on the image. Use windowing (ch 4) and
steps for converting a color image to gray scale (as in 3.7.4) to emphasize the spots in the image.
Please don’t loop through the image as the author does in 3.7.4. Make sure you choose settings
that maximize the contrast between these spots and everything else in the image.
(a) (1 point) Describe how you chose your windowing parameters and display the parameters.
(b) (1 point) Describe why you chose the weights you used to create the grayscale image.
(c) (2 points) Display the final result.
(5) This exercise builds off Practical Lesson 5.4.3. The author has supplied code for computing
the forward difference in the x direction in a file called, ForwardDifference_5.m.
(a) (3 points) Recreate the left panel of Figure 5.7. Also, figure out how to properly use the
MATLAB function, conv2 to carry out the convolution. Display the kernel you used and
make a 2 panel plot with the author’s result in the left panel and conv2 in the right panel.
Include appropriate figure headings. Be careful! Please refer back to the class notes on the
proper definition of convolution.
(b) (1 point) Repeat the previous part, but compute the derivative in the y direction. Again, use
both the author’s looping strategy and conv2. Display the kernel you used as well as the 2
panel plot of the author’s result and conv2 result side-by-side.
(c) (2 points) Compute the length of the gradient (see equation on p. 123 of the book) to estimate
the edges of the image. Note, I’d like you to use the central difference kernel, but make sure
your kernel definition matches the definition where x corresponds to the rows of the matrix
and y the columns (similar to what you did in part a). You may use the conv2 operation.
Display the result as well as your two kernels.