Frank Delaglio, Ph.D.

19804 Maycrest Way
Germantown MD 20876 USA

301 806-0867
delaglio@nmrscience.com



NMRPipe Processing Functions
IMG: Image Processing Utilities.

Flag Argument Default Description
 -conv Apply Convolution Filter (Default). Requires -kern ...
Statistical Filters and Options:
 -median Median.
 -min Minimum (Erosion).
 -max Maximum (Dilation).
 -amin Minimum Absolute Value.
 -amax Maximum Absolute Value.
 -range Range.
 -avg Average.
 -dev Standard Deviation.
 -thresh t 0.0 Use Threshold t for Selection.
General Filter Parameters:
 -dx dx 1 +/- Filter Width, Points.
 -dy dy 1 +/- Filter Height, Points.
Convolution Filter Parameters:
 -kern kList List of Kernel Values.

IMG provides some simple image processing filters primarily intended for 2D planes. Given a 2D plane Zi,j, a given filter creates a new 2D plane Z'i,j where a given point Z'i=x,j=y is a function F( ) of all points in a rectangular region of the original 2D plane, where the region sizes are specified in terms of +/- points dx and dy:

      Z'i=x,j=y = F( Zk,m ) where k = x - dx to x + dx, and m = y - dy to y + dy.

Regions which extend past the edges of the image are wrapped to the opposite sides. The function applied is specified according to the following flags:

      Flag Function F( )
       -conv Convolution (Default)
       -median Median
       -min Minimum (Erosion).
       -max Maximum (Dilation).
       -amin Minimum Absolute Value.
       -amax Maximum Absolute Value.
       -range Range.
       -avg Average.
       -dev Standard Deviation.

OPTIONS

-kern cList
This option is used to specify the coefficients Ck,m used in the case where a convolution in applied:

      Z'i=x,j=y = Ck,mZi+k-dx,j+m-dy where k = 0 to 2 dx, and m = 0 to 2 dy.

-dx dx
Specifies dx, the +/- X-Axis size of the image processing neighborhood, in points.

-dy dy
Specifies dy, the +/- Y-Axis size of the image processing neighborhood, in points.

-thresh t
If a threshold option is given, only points in the original image which are greater than the specified will be used for computing F().

EXAMPLES

The following is an example of a convolution filter which corresponds to a simple form of image sharpening. The coefficients of the 3x3 convolution kernel are listed on multiple lines to emphasize their two-dimensional structure, although this is not required:

   nmrPipe -in image.dat \
   | nmrPipe -fn IMG -dx 1 -dy 1 -conv \
             -kern   -1 -1 -1 \
                     -1  9 -1 \
                     -1 -1 -1 \
     -out sharp.dat -ov