A short quiz on intermediate-level vision

Adrian F. Clark

1. Why does the first stage of Canny involve convolution with a blurring mask?

   • To reduce the number of pixels

   • To reduce the effects of noise

   • To perform non-maximum suppression
   • To detect edges

   Blurring smooths out an image, reducing the effects of noise.

2. How are the differences between pixels processed in Canny?

   • Changes in direction indicate corners
   • They are passed into a non-maximum suppression algorithm

   • They are converted to magnitude and direction

   • They are converted into edges

   The pixel differences are of little direct use; they are converted into magnitude and orientation, which are much easier to work with.

3. The first stage of the Canny edge detector involves smoothing with what?

   • a Gaussian

   • a Laplacian mask
   • a Sobel mask
   • a 3 x 3 blur mask

   The first stage of the Canny edge detector is to smooth the image (in the hope of reducing the number of false edges) with a Gaussian mask.

4. SURF is a speeded-up version of which feature detector and descriptor?

   • ORB
   • FAST
   • BRIEF

   • SIFT

   SURF is modelled on SIFT but uses operators that are intended to be faster to compute.

5. What does one do with the Hough accumulator to find lines?

   • Each corner in the accumulator is a line in the image

   • Each peak in the accumulator is a line in the image

   • Each 3 x 3 region in the accumulator is a line in the image
   • Each line in the accumulator is a line in the image

   As the code in the notes illustrates clearly at the end of Chapter 5, peaks in the Hough accumulator correspond to lines in the image. This is because more image pixels have `voted' for that set of parameters than any other set.

6. How does one perform non-maximum suppression?

   • Take the largest-valued edge magnitude parallel to the edge direction
   • Take the largest-valued edge orientation parallel to the largest edge magnitude

   • Take the largest-valued edge magnitude perpendicular to the edge direction

   • Take the largest-valued edge orientation perpendicular to the largest edge magnitude

   We want to find the largest value of the edge magnitude across its profile, and that means we have to look perpendicular to the direction in which the edge lies.

7. Which of the following can a Hough transform not be used to identify?

   • Faces

   • Ellipses
   • Circles
   • Straight lines

   Alas, the Hough transform cannot be used to identify faces because a face does not have a (small enough) set of parameters that describe it in an image.

8. Which of the following is suitable for matching SIFT features P and Q?

   • the expression sqrt{ (P(x,y) - Q(x, y))^2}

   • the expression P(x,y) x Q(x, y)
   • the expression sqrt{ P(x,y) } - sqrt{ Q(x, y) }
   • the expression sqrt{ P(x,y)^2 + Q(x, y)^2 }

   We take the difference between features in $P$ and $Q$. As this will contain both positive and negative values and we want to know how much they differ overall, we need to change negative values into positive ones; we normally do this by squaring this difference.

9. How are BRIEF features matched?

   • exclusive OR

   • OR
   • Euclidean distance
   • angle between vectors

   BRIEF is intended to run in real time and uses strings of binary digits as the descriptor. The easiest way to measure dissimilarity of two bit-strings is to use exclusive-OR and then count the number of set bits.

10. How should peaks chosen from the Hough accumulator array?

    • Nearest to origin first

    • Largest first

    • Smallest radius first
    • Smallest first

    The peak with the most lines in the accumulator contributing to it is the most significant line, and this is the pixel with the largest value.