Skip to content

RS_Resample

Resamples a raster onto a new pixel grid — a new width/height, a new pixel size, and/or a snapped origin — in the same CRS.

Usage

raster RS_Resample(raster: raster, reference_raster: raster, use_scale: boolean, algorithm: string)
raster RS_Resample(raster: raster, width_or_scale: double, height_or_scale: double, use_scale: boolean, algorithm: string)
raster RS_Resample(raster: raster, width_or_scale: double, height_or_scale: double, grid_x: double, grid_y: double, use_scale: boolean, algorithm: string)

Arguments

  • raster (raster): The input raster to resample.
  • width_or_scale (double): Output width in pixels when use_scale is false (a fractional value is an error), or the target x pixel size (scaleX) when use_scale is true.
  • height_or_scale (double): Output height in pixels when use_scale is false (a fractional value is an error), or the target y pixel size (scaleY, negative for a north-up raster) when use_scale is true.
  • grid_x (double): X coordinate of the grid the output origin snaps to.
  • grid_y (double): Y coordinate of the grid the output origin snaps to.
  • use_scale (boolean): When false, width_or_scale/height_or_scale are output pixel dimensions and the world extent is preserved; when true, they are the target pixel size.
  • algorithm (string): Resampling algorithm (case-insensitive): NearestNeighbor (the default), Bilinear, Cubic (alias Bicubic), CubicSpline, Lanczos, Average, or Mode.
  • reference_raster (raster): A raster whose grid — dimensions (or pixel size, when use_scale is true) and origin — defines the output grid. Must share the input’s CRS.

[!WARNING]

Experimental. This function is experimental; its behavior may change without notice.

Description

RS_Resample changes a raster’s pixel grid while keeping its band count and order and per-band nodata. Pixel values are recomputed by GDAL with the chosen algorithm (nearest neighbour by default). The output stays in the input’s CRS — RS_Resample never reprojects (use RS_ReprojectMatch for that).

RS_Resample has three positional overloads:

  • RS_Resample(raster, width_or_scale, height_or_scale, use_scale, algorithm) — when use_scale is false, width_or_scale/height_or_scale are the output pixel dimensions (a fractional value is an error) and the world extent is preserved. When use_scale is true they are the target pixel size (scaleX, scaleY, with scaleY negative for a north-up raster); the pixel size is kept exact and the extent grows by up to one pixel so it tiles into whole pixels. Cells outside the source footprint become nodata.
  • RS_Resample(raster, width_or_scale, height_or_scale, grid_x, grid_y, use_scale, algorithm) — additionally snaps the output origin to the grid anchored at (grid_x, grid_y), expanding outward so the original extent stays covered.
  • RS_Resample(raster, reference_raster, use_scale, algorithm) — takes the target dimensions (or pixel size, when use_scale) and the origin grid from reference_raster, which must share the input’s CRS (it is an error otherwise).

algorithm (case-insensitive) is one of NearestNeighbor (the default), Bilinear, Cubic (alias Bicubic), CubicSpline, Lanczos, Average, or Mode.

Dimensions and pixel size use the rows-then-columns intuition of the raster (height_or_scale/scaleY for rows, width_or_scale/scaleX for columns). In the extent-preserving dimension form, a skewed or rotated raster is re-gridded with its footprint preserved (both the scale and skew terms of each axis scale by the same factor). A scale change or origin snap on a skewed raster is not supported and errors.

For an N-D raster (bands with extra dimensions such as time), the resample is a 2-D (y, x) operation broadcast across every non-spatial plane, so the non-spatial dimensions are preserved and only the (y, x) extent changes.

[!NOTE]

Extent-growing scale changes and origin snaps read the entire raster into memory.

Examples

Downsample the example raster (64 x 32) to half its dimensions:

SELECT RS_Width(RS_Resample(RS_Example(), 32, 16, false, 'NearestNeighbor'));

┌──────────────────────────────────────────────────────────────────────────────┐
│ rs_width(rs_resample(rs_example(),Int64(32),Int64(16),Boolean(false),Utf8("N │
│                              earestNeighbor")))…                             │
╞══════════════════════════════════════════════════════════════════════════════╡
│                                                                           32 │
└──────────────────────────────────────────────────────────────────────────────┘

Upsample to twice its dimensions with bilinear resampling:

SELECT RS_Height(RS_Resample(RS_Example(), 128, 64, false, 'Bilinear'));

┌──────────────────────────────────────────────────────────────────────────────┐
│ rs_height(rs_resample(rs_example(),Int64(128),Int64(64),Boolean(false),Utf8( │
│                                "Bilinear")))…                                │
╞══════════════════════════════════════════════════════════════════════════════╡
│                                                                           64 │
└──────────────────────────────────────────────────────────────────────────────┘