image2lz {fields}R Documentation

Some simple functions for subsetting images


These function help in subsetting images or reducing its size by averaging adjecent cells.


crop.image(obj, loc=NULL,...)
average.image(obj, Q=2)
in.poly( xd, xp, convex.hull=FALSE, inflation=1e-07)
in.poly.grid( grid.list,xp, convex.hull=FALSE, inflation=1e-07)



A list in image format with the usual x,y defining the grid and z a matrix of image values.


A 2 column matrix of locations within the image region that define the subset. If not specified then the image is plotted and the rectangle can be specified interactively.


Number of pixels to average.


A 2 column matrix of locations that are the points to check for being inside a polygon.


A 2 column matrix of locations that are vertices of a polygon. The last point is assumed to be connected to the first.


If TRUE then the convex hull of xp is used instead of the polygon.


A list with components x and y specifing the 2-d grid values. (See help( grid.list) for more details.)


A small expansion factor to insure that points precisely on the boundaries and vertices of the convex hull are included as members.


Graphics arguments passed to image.plot. This is only relevant when loc is NULL and the locator function is called via get.rectangle.


If loc has more than 2 rows then the largest rectangle containing the locations is used.


Creates a subset of the image obj by taking using the largest rectangle in the locations loc. This is useful if one needs to extract a image that is no bigger in extant than som edata location. If locations are omitted the parent image is plotted and the locations from two mouse clicks on the image. Returned value is an image with appropriate x,y and z components.


Given an image plots and waits for two mouse clicks that are returned.

average.image, half.image

Takes passed image and averages the pixel values and adjusts the grid to create an image that has a smaller number of elements. If Q=2 in average.image it has the same effect as half.image but might be slower – if the original image is mXn then half image will be an image (m/2)X(n/2). This begs the question what happens when m or n is odd or when (m/Q) or (n/Q) are not integers. In either case the largest rows or columns are dropped. (For large Q the function might be modified to drop about half the pixels at both edges.)

in.poly, in.poly.grid

Determines whether the points xd,yd are inside a polygon or outside. Return value is a logical vector with TRUE being inside or on boundary of polygon. The test expands the polygon slightly in size (on the order of single precision zero) to include points that are at the vertices. in.poly does not really depend on an image format however the grid version in.poly.grid is more efficient for considering the locations on a regular grid See also that is hard coded to work with the fields world map.


Doug Nychka

See Also

drape.plot, image.plot, interp.surface, interp.surface.grid,



# region defining Colorado Front Range

  loc<- rbind( c(-106.5, 40.8),
             c(-103.9, 37.5))

# extract elevations for just CO frontrange.
   FR<- crop.image(RMelevation, loc)
   image.plot( FR, col=terrain.colors(256))

# average cells  4 to 1 by doing this twice!
   temp<-  half.image( RMelevation)
   temp<- half.image( temp)

# or in one step
   temp<-  average.image( RMelevation, Q=4)-> temp
   image.plot( temp, col=terrain.colors(256))

# a polygon (no special meaning entered with just locator)
x1p<- c(
 -106.2017, -104.2418, -102.9182, -102.8163, -102.8927, -103.3254, -104.7763,
 -106.5581, -108.2889, -109.1035, -109.3325, -108.7980)

x2p<- c(
  43.02978, 42.80732, 41.89727, 40.84566, 39.81427, 38.17618, 36.53810, 36.29542,
  36.90211, 38.29752, 39.45025, 41.02767)
xp<- cbind( x1p,x2p)

 image.plot( temp)
 polygon( xp[,1], xp[,2], lwd=2)

# find all grid points inside poly
 fullset<- make.surface.grid( list( x= temp$x, y= temp$y))
 ind<-  in.poly( fullset,xp)

# take a look 
 plot( fullset, pch=".")
 polygon( xp[,1], xp[,2], lwd=2)
 points( fullset[ind,], pch="o", col="red", cex=.5)

# masking out the image NA == white in the image plot
 temp$z[!ind] <- NA
 image.plot( temp)
 polygon( xp[,1], xp[,2], lwd=2)

# This is more efficient for large grids:
# because the large number of grid location ( xg above) is 
# never explicitly created.

 ind<- in.poly.grid( list( x= temp$x, y= temp$y), xp)

# now use ind in the same way as above to mask points outside of polygon

[Package fields version 8.4-1 Index]