6

I am currently working in SIFT, I had generated the difference of Gaussian and the extrema image layers. Can anyone explain to me how to use Hessian matrix to eliminate the low contrast keypoint?

Amir
  • 10,600
  • 9
  • 48
  • 75
tsann
  • 89
  • 4
  • 1
    You might want to explain yourself a little better. Not many people specialize in CV. Also, the correct tagging of your question helps others find your question better. – monksy Apr 22 '12 at 00:10
  • I am currently working on image feature extration to form descriptor for my pattern matching. – tsann Apr 22 '12 at 00:28
  • I found this post really clear in explaining and implementation about SIFT http://www.aishack.in/2010/05/sift-scale-invariant-feature-transform/ And maybe this one will help you: http://www.aishack.in/2010/05/sift-scale-invariant-feature-transform/5/ – vancexu Aug 01 '12 at 03:34
  • The Hessian matrix is used to eliminate features along edges/lines not `low contrast keypoints`. See the [related section of the SIFT entry](http://en.wikipedia.org/wiki/Scale-invariant_feature_transform#Eliminating_edge_responses) on Wikipedia. This is a rather simple operation. – fireant May 03 '12 at 17:15
  • The above links doesn't work. There's now: http://aishack.in/tutorials/sift-scale-invariant-feature-transform-introduction/ and http://aishack.in/tutorials/sift-scale-invariant-feature-transform-eliminate-low-contrast/ – Utkarsh Sinha Aug 31 '14 at 05:13

1 Answers1

1

A good keypoint is a corner. This comes from the Harris corner work and the Good features to track (KLT) papers first, then emphasized by the Mikolajczyk and Schmid paper.

Intuitively, a corner is a good feature because it is an intersection of two lines, while a single line segment can be moved along its direction, thus causing a less accurate localization. A line segment is an edge, i.e., a first order derivative (gradient). A corner is an edge that changes its direction abruptly. This is measured by a second order derivative, hence the use of the Hessian matrix that contains the values of the directional second derivatives.

sansuiso
  • 9,259
  • 1
  • 40
  • 58