Brain Imaging & Analysis Center

usage: vdst = filter(vsrc,ftype,fsize,offset,minval,maxval,shift,vmask,dmax)

FILTER is a form of convolution wherein each pixel in an image is replaced by the result of a function performed on that pixel and its 8 immediate neighbors. This is different from the CONVOLVE(3) function in that the convolution function is not specified by a kernel matrix. Instead, the function is selected from a predefined list using the FTYPE argument.

Legal FTYPE values are:

     0   MAX             Replace with the maximum value of its neighbors
     1   MIN             Replace with the minimum value of its neighbors
     2   ADAPT
     3   MEAN            Replace with the mean of its neighbors
     4   STDEV           Replace with the standard deviation of its neighbors
     5   MEDIAN          Replace with the median of its neighbors
     6   ERODE      (b)  Thin objects by 1 pixel on each side
     7   DILATE     (b)  Expand objects by 1 pixel on each side
     8   SOBEL           Perform a Sobel edge enhancement:
                             a b c          x = (a+2c+c)-(g+2h+i)
                             d e f   ->     y = (a+2d+g)-(c+2f+i)
                             g h i       e' = sqrt(x**2 + y**2)
     9   CONNECT    (b)
    10   NEIGHBORS  (b)
    11   FILLIN          Replace zero values with the mean of its non-zero
                         neighbors
    12   SMOOTH          Replace each value with the mean of those neighbors
                         that differ by no more than DMAX from the original
                         center value (DMAX default is 10)

 (b) indicates filters that treat VSRC as binary data, either 0 or 1-- in which
 case any data pixel value greater than MAXVAL is treated as a 1.

See Also:
CONVOLVE(3), SMOOTH(3)

CIGAL Home, CIGAL Manual, Functions List, Manual Help