------------------------------------------------------------------
Please write your answer below each question
------------------------------------------------------------------
------------------------------------------------------------------
Part B.1 Questions (you must answer questions 1 & 2 to
get any credit for Part B.1!!)
------------------------------------------------------------------
1. Why did the inpainting algorithm work well on your first
dataset (with Source1, Mask1 as input)?
2. Why is your second dataset (Source2, Mask2) a "bad" case for
inpaiting? Specifically why do you expect the algorithm to
perform poorly on the Source2 image for EVERY setting of the
patch radius?
------------------------------------------------------------------
Part B.2 Questions
------------------------------------------------------------------
Question 1 (5 points)
The top-level loop of the inpainting algorithm is implemented by
as a while loop in method
inpainting::compute() in file src/inpainting/inpainting_algorithm.cxx
Even though this loop causes the implementation to behave exactly
"as advertised" in the paper for most cases of source images and
initial masks, there are situations where this is *not* the case.
In such cases, the implementation is expected to generate inpainted
images that have significant artifacts.
You need to identify what the problem is with this particular loop.
[Hint: To answer this question, you do *not* need to look at the code
of the inpaint_region() and find_first_unfilled_pixel() methods
or of the vil_trace_8con_boundary() function (although you do need
to understand what these methods do). Also, the answer
has nothing to do with variables iterations_done, max_iterations,
and inpainted_new_region]
------------------------------------------------------------------
Question 2 (15 points)
Prove that if gamma(t) = ( x(t), y(t) ) is a 2D curve, the curvature
k(t) at point gamma(t) is given by the expression
2 2
dx d y d x dy
-- * --- - --- * --
2 2
dt dt dt dt
k(t) = -------------------------
- - 3/2
| dx 2 dy 2 |
| ( -- ) + ( -- ) |
| dt dt |
- -
(This question is identical to the one on the Feb 5th lecture slides)
------------------------------------------------------------------