Harris Corner Detector

Question 1

Harris corner detector process

Answer 1

Compute partial derivatives of image

Compute product of derivatives at every pixel

Compute the weighted sums of derivatives using a Gaussian window around each pixel

Define second moment matrix M at each pixel

Compute the corner response function at each pixel

Threshold value of R
find local maxima (non-max suppression)

Question 2

What does changing the scale of the image do when using the Harris corner detector

Answer 2

Corner location is not covariant, meaning the results will change if you scale the image

Question 3

Scale invariant

Answer 3

Design a function that when u take the local maximum, can be used as an anchor to calculate how the image scaled and the new correlating pixels

Signature function

Question 4

What makes a good signature function?

Answer 4

Has one stable sharp peak

Sharp local intensity change for usual images

Question 5

Laura Ian of Gaussian can also be used to detect

Answer 5

Blobs with images, the blobs have to line up with the (Mexican hat) like protrusion within the filter