As with any in-class exam, you are encouraged to ask questions if you need clarification or are stuck. You can post your question here or e-mail me.

If my answer contains important hints that could help others, I will post this answer here for the benefit of the entire class.

The exam is due 12-08-16 at 5:00 pm.

Hey Dr. Haidekker,

For question number 4 would you like us to explain the reasoning behind why our kernel works? I kind of figured it out by creating different kernels in a program and studying the way changing values changed certain aspects of the picture but once I figured it out, I understand why it works. Would you like us to explain that?

ANDDD, I just read the directions again and realized that, yes, I should add explanation. Sorry again!

Some explanation never hurts. If you got it right, it is completely obvious that you understood it. If you did NOT get it right, however, your reasoning MAY get you some partial credit.

Also from number 5 is a calibration wedge just something used to calibrate an x-ray machine?

I feel like this question was really stupid and obviously, yes, that is what it is! Haha sorry Dr. H!!

Not a stupid question — it may not be immediately obvious. It is not critically important what modality you are dealing with (Lambert-Beer’s law gives it away, I guess), but perhaps it would help understanding where this question aims?

yes it does, it doesn’t matter for the question, I was just curious!

For question 13, when you say multiple closing operations are not a suitable answer does that mean that we shouldn’t use erosion or dilation at all?

Also, I’m guessing that also means we can’t just use image value? hehe

What would $n$ dilations followed by $n$ erosions do? Here, $n$ is the number of missing pixels.

Also, could you imagine that the first part of the bonus problem sort of prepares for the second?

ohh okay! One more question!

for the first part of this question. Do you mean a skeleton like this:

http://rdbms-insight.com/wp/wp-content/uploads/2014/10/antique_anatomy_illustration__human_skeleton_circa_1911_sjpg3242.jpg

or the second part of this image:

https://www.researchgate.net/profile/A_Kolesov/publication/253034195/figure/fig8/AS:298094680920071@1448082718330/Figure-1-a-the-discrete-binary-image-b-the-discrete-skeleton-c-the.png

The second option.

thank you!

Clarification question: on question 12b.

“There is a large spread of shape irregularity.”

“large spread” = high?

Does this mean shape irregularity represents a high value, or does this mean there are spread out values all along the ‘shape irregularity’ axis?

should A, B, C be represented by single points in our 3d space?

Thanks,

“Should A, B, C be represented by single points in our 3d space?”Possible, but I was thinking about hinting at the locations as a cloud, a hashed region, or similar. This is entirely qualitative.

â€ślarge spread = high?”No. Rather, there is large variability from low to high.

I have a question about 6, well, several questions.

For part a:

I’m unsure how detailed you want the equation to be. Would you like the equation completely simplified down to solving the integrations?

For part b: Again, would you like integrals or is it enough to mention the parts of the equation and their fourier transforms? also, in the Hints section, you ask us to use G= F{g} and N = F{n} what do those pairs correlate to exactly? Can we assign to them whichever markers we want for them to be in our equation?

Thank you!

None of the information you are asking for is given. This is a purely conceptual question. To give you an example, the degradation in the Fourier domain can be written as

$$

I(u,v) = I_{ID}(u,v) \cdot G(u,v) + N(u,v)

$$

This is the level of abstraction for this question.

Okay, I was just making sure! Thank you!

For question 5 b, what should we look for in an “idealized expected value”? I’m struggling to find out what this is referring to.

The problem hints at Lambert-Beer’s law. You have enough data points to determine the unknowns. This also means that you can calculate the intensity underneath the third step. OK so far, I guess.

Now you need to store this in an 8-bit pixel of the image. Can you? If not, what happens to the value?