landsat (version 1.1.2)

georef: Simple image-matching georeferencing function.

Description

Finds best fit between target image and tofix image by minimizing RMSE between the two. The tofix image is moved one pixel at a time horizontally or vertically. Simple automated georeferencing is adequate for some image-processing tasks.

Usage

georef(target, tofix, maxdist = 1000, startx = 0, starty = 0)

Value

shiftx

The x-direction shift to get the best match (lowest RMSE).

shifty

The y-direction shift to get the best match (lowest RMSE).

initrmse

Initial RMSE between target and tofix images.

currrmse

Lowest RMSE, after shiftx and shifty pixel adjustments. Will be 9999 if maxdist is exceeded.

Arguments

target

A georeferenced base image; can be matrix, dataframe or SpatialGridDataFrame.

tofix

The image to be georeferenced; can be matrix, dataframe or SpatialGridDataFrame.

maxdist

The greatest distance to move the tofix image. If this is exceeded, the function will stop.

startx

Shift the tofix image this many pixels in the x direction before beginning, to avoid local minimum.

starty

Shift the tofix image this many pixels in the y direction before beginning, to avoid local minimum.

Author

Sarah Goslee

Details

This function offers a simplistic approach to georeferencing using an iterative algorithm that at each step moves the tofix image one pixel in the direction that produces the greatest reduction in RMSE. When RMSE no longer decreases or maxdist is reached, the algorithm stops, assuming that the tofix image now matches the reference target image. This algorithm can produce local minima. Results should always be checked visually.

Note: this algorithm is only effective with images larger than the samples included with this package. The July and November images are already georectified, but this function will show them as needing considerable adjustment. Images of at least 1000x1000 pixels are necessary for adequate results.

See Also

geoshift

Examples

# to use for georeferencing
data(nov3)
data(july3)
july.shift <- georef(nov3, july3, maxdist=50) # match july to november
july3.corr <- geoshift(july3, padx=50, pady=50, july.shift$shiftx, july.shift$shifty)

# only need to run georef once for a particular date
# use the same correction for all bands
data(july4)
july4.corr <- geoshift(july4, padx=50, pady=50, july.shift$shiftx, july.shift$shifty)