Kirsch operator

Edge detector

The Kirsch operator or Kirsch compass kernel is a non-linear edge detector that finds the maximum edge strength in a few predetermined directions. It is named after the computer scientist Russell Kirsch.

Mathematical description

The operator takes a single kernel mask and rotates it in 45 degree increments through all 8 compass directions: N, NW, W, SW, S, SE, E, and NE. The edge magnitude of the Kirsch operator is calculated as the maximum magnitude across all directions:

h n , m = max z = 1 , , 8 i = 1 1 j = 1 1 g i j ( z ) f n + i , m + j {\displaystyle h_{n,m}=\max _{z=1,\dots ,8}\sum _{i=-1}^{1}\sum _{j=-1}^{1}g_{ij}^{(z)}\cdot f_{n+i,m+j}}

where z enumerates the compass direction kernels g:

g ( 1 ) = [ + 5 + 5 + 5 3 0 3 3 3 3 ] ,   {\displaystyle \mathbf {g^{(1)}} ={\begin{bmatrix}+5&+5&+5\\-3&0&-3\\-3&-3&-3\end{bmatrix}},\ } g ( 2 ) = [ + 5 + 5 3 + 5 0 3 3 3 3 ] ,   {\displaystyle \mathbf {g^{(2)}} ={\begin{bmatrix}+5&+5&-3\\+5&0&-3\\-3&-3&-3\end{bmatrix}},\ } g ( 3 ) = [ + 5 3 3 + 5 0 3 + 5 3 3 ] ,   {\displaystyle \mathbf {g^{(3)}} ={\begin{bmatrix}+5&-3&-3\\+5&0&-3\\+5&-3&-3\end{bmatrix}},\ } g ( 4 ) = [ 3 3 3 + 5 0 3 + 5 + 5 3 ] {\displaystyle \mathbf {g^{(4)}} ={\begin{bmatrix}-3&-3&-3\\+5&0&-3\\+5&+5&-3\end{bmatrix}}} and so on.

The edge direction is defined by the mask that produces the maximum edge magnitude.

Example images

  • Original
    Original
  • Maximum gradient in the 8 directions
    Maximum gradient in the 8 directions
  • Image filtered with g(1)
    Image filtered with g(1)
  • Image filtered with g(2)
    Image filtered with g(2)
  • Image filtered with g(3)
    Image filtered with g(3)
  • Image filtered with g(4)
    Image filtered with g(4)
  • Image filtered with g(5)
    Image filtered with g(5)
  • Image filtered with g(6)
    Image filtered with g(6)
  • Image filtered with g(7)
    Image filtered with g(7)
  • Image filtered with g(8)
    Image filtered with g(8)

References

  • Kirsch, R. (1971). "Computer determination of the constituent structure of biological images". Computers and Biomedical Research. 4 (3): 315–328. CiteSeerX 10.1.1.161.956. doi:10.1016/0010-4809(71)90034-6.