How to crop a raster image by coordinates in python?

Question

I have a GeoTiff in UTM32 and coordinates of a rectangle also in UTM32. (This projection may not always be the case, but the projections will always be the same)

I simply need to crop the image using the rectangle .

The rectangle is given by: (xmin, xmax, ymin, ymax)

699934.584491, 700160.946739, 6168703.00544, 6169364.0093

I know how to make a polygon from the points, how to make a shapefile from the polygon, and I know how to create a masked numpy array using the points. However, I don't know how to use the polygon, the shapefile or the mask to actually crop the image.

I already looked at the description at: https://pcjericks.github.io/py-gdalogr-cookbook/raster_layers.html#clip-a-geotiff-with-shapefile

However, I don't really understand it and it seems overly complicated. (like I don't know what histogram stretching is supposed to be doing there except confusing)

Answer 1

Try to use bbox = (xmin,ymin,xmax,ymax)

from osgeo import gdal

bbox = (xmin,ymin,xmax,ymax)

gdal.Translate('output_crop_raster.tif', 'input_raster.tif', projWin = bbox)

Answer 2

Akin's answer is right, but doesn't provide a complete explanation.

You can crop a gdal file using gdal_translate , which can be used in python via gdal.Translate .

Best option: projwin

The easiest way is with the projwin flag, which takes 4 values:

window = (upper_left_x, upper_left_y, lower_right_x, lower_right_y)

These values are in map coordinates. The bounds of the input file can be obtained via gdalinfo input_raster.tif from the command line.

NOTE : for many coordinate systems, ymax is actually less than ymin , so it's important to use "upper_left" and "lower_right" to identify the coordinates instead of "max" and "min."

The complete solution, then is:

from osgeo import gdal

upper_left_x = 699934.584491
upper_left_y = 6169364.0093
lower_right_x = 700160.946739
lower_right_y = 6168703.00544
window = (upper_left_x,upper_left_y,lower_right_x,lower_right_y)

gdal.Translate('output_crop_raster.tif', 'input_raster.tif', projWin = window)

Additional option: srcwin

srcwin is another gdal_translate flag similar to projwin , but takes in the pixel and line window via an offset and size, instead of using the map coordinate bounds. You would use it like this:

window = (offset_x, offset_y, size_x, size_y)
gdal.Translate('output_crop_raster.tif', 'input_raster.tif', srcWin = window)