How can I find the cells which lie on a straight line between two points? I.e. how can I find the indices of the cells (in grey) which are intersected by the line defined by the two points (in blue):

enter image description here

I want to do this in order to read off values from the underlying DTM and thus extract the height profile along the line....

Using the geotransform of the raster, it is easy to convert the two points to cell offsets, so my first thought was to create many points along the line and convert each of these. However the problem here is creating points at the right distance to successfully capture each intersected cell (while also not creating a ridiculous number of points so that it would be incredibly inefficient)

Looking for ideas that are either programmatic (python/C++) or theory (e.g. no arcgis tools! gdal ok...)

2 Answers 2


You can use gdal_rasterize command as base of the procedure. I tested my approach with this situation:

enter image description here

The code is:

import os

#extent of base raster
extent = "-te 354971.3488602247089148 4473058.4587739501148462  355341.2924763654009439 4473428.4023900907486677 "
attribute_name = " -a id "
#resolution of base raster
resolution = " -tr 73.9887 73.9887 "
input_vector_line = " -l test_line /home/zeito/pyqgis_data/test_line.shp "
output_raster = " /home/zeito/pyqgis_data/test_line_raster.tif"

cmd = "gdal_rasterize -at " + \
                     extent + \
             attribute_name + \
                 resolution + \
          input_vector_line + \


where the option -at is very important because "Enables the ALL_TOUCHED rasterization option so that all pixels touched by lines or polygons will be updated, not just those on the line render path, or whose center point is within the polygon. Defaults to disabled for normal rendering rules" (http://www.gdal.org/gdal_rasterize.html).

After running the code at the Python Console of QGIS I got:

enter image description here

where the line was adequately rasterized with a constant value of 1. Now, with map algebra, you can extract the height profile along the line.

You can use the GDAL options to get raster extent, resolution, etc, for automatizing all procedure.


The idiomatic algorithm for extraction of pixels closest to a line is the Bresenham algorithm. It may be present in underlying guts of libraries such as GDAL or others.

You will most probably find lots of implementations in various languages with some web search.

Quoting wikipedia:

Bresenham's line algorithm is a line drawing algorithm that determines the points of an n-dimensional raster that should be selected in order to form a close approximation to a straight line between two points [emphasis mine]. It is commonly used to draw line primitives in a bitmap image (e.g. on a computer screen), as it uses only integer addition, subtraction and bit shifting, all of which are very cheap operations in standard computer architectures. It is an incremental error algorithm. It is one of the earliest algorithms developed in the field of computer graphics [emphasis mine].

That being said, i discovered a few days ago this approach that looks quite elegent using GDAL. You can find the code on the author's github.

It uses pyproj projections + GDAL's Warp feature, in order to obtain a 1D array of elevations along the geodesic of 2 inputs positions.

The code is MIT licensed, but for the sake of stackexchange phisolophy, I'll paste it here. This is not mine, but kokoalberti's :

    Simple utility program to extract profiles from GDAL data sources using a 
    two-point equidistanct projection.
    - Add option to include coordinates in output csv file. Currently it just
            records along-profile distance.
    import argparse
    import sys
    from pyproj import CRS, Transformer
    from osgeo import gdal
    if __name__ == '__main__':
        # Parse command line args
        parser = argparse.ArgumentParser(description="Make profile with GDAL")
        parser.add_argument('src', metavar='SRC', help='GDAL data source')
        parser.add_argument('lon_1', metavar='LON_1', type=float, help='Longitude of start point')
        parser.add_argument('lat_1', metavar='LAT_1', type=float, help='Latitude of start point')
        parser.add_argument('lon_2', metavar='LON_2', type=float, help='Longitude of end point')
        parser.add_argument('lat_2', metavar='LAT_2', type=float, help='Latitude of end point')
        parser.add_argument('--width', type=int, default=100, help='Profile width (m)')
        parser.add_argument('--dist', type=int, default=100, help='Profile sampling distance (m)')
        parser.add_argument('--resample', default='near', help='Resampling method')
        parser.add_argument('--tif', default='', help='Output GTiff file')
        parser.add_argument('--csv', default='', help='Output CSV file')
        args = parser.parse_args()
        # Open/validate the source dataset
        ds = gdal.Open(args.src)
        if not ds:
            print("Could not open dataset.")
        # Set up coordinate transform
        proj_str = "+proj=tpeqd +lon_1={} +lat_1={} +lon_2={} +lat_2={}".format(args.lon_1, args.lat_1, args.lon_2, args.lat_2)
        tpeqd = CRS.from_proj4(proj_str)
        transformer = Transformer.from_crs(CRS.from_proj4("+proj=latlon"), tpeqd)
        # Transfor to tpeqd coordinates
        point_1 = transformer.transform(args.lon_1, args.lat_1)
        point_2 = transformer.transform(args.lon_2, args.lat_2)
        # Create an bounding box (minx, miny, maxx, maxy) in tpeqd coordinates
        bbox = (point_1[0], -(args.width*0.5), point_2[0], (args.width*0.5))
        # Calculate the number of samples in our profile.
        num_samples = int((point_2[0] - point_1[0]) / args.dist)
        print("Reading and warping data into profile swath...")
        # Warp it into dataset in tpeqd projection. If args.tif is empty GDAL will
        # interpret it as an in-memory dataset.
        format = 'GTiff' if args.tif else 'VRT'
        profile = gdal.Warp(args.tif, ds, dstSRS=proj_str, outputBounds=bbox, 
                            height=1, width=num_samples, resampleAlg=args.resample, 
        # Extract the pixel values and write to an output file
        data = profile.GetRasterBand(1).ReadAsArray()
        print("Created {}m profile with {} samples.".format(args.dist*num_samples, num_samples))
        # Write csv output
        if args.csv:
            with open(args.csv, 'w') as f:
                for (d, value) in enumerate(data[0,:]):
                    f.write("{},{}\n".format(d*args.dist, value))
            print("Saved as {}".format(args.csv))
        # Clean up
        profile = None
        ds = None

Last link which I do not know if it is worth it (paid content ?) on oreilly seems to rely on GDAL and Shapely to achieve same result, if someone has some feedback on this, feel free.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.