Module resample 

Resampling kernels — the math that makes pyramid building and raster warp possible.

API shape

Two layers of API:

• Point sampling (sample_*): given a band + fractional (x, y) in source pixel space, return the resampled value. Used by warp to map one output pixel to its source value.
• Tile-to-tile (resample_tile): given a whole source RasterTile + target dimensions, return a new RasterTile resampled across all bands. Used by pyramid to downsample for overview levels.

Kernels covered

• ResamplingMethod::Nearest — fastest; preserves categorical values exactly. Use for landcover/classification.
• ResamplingMethod::Bilinear — 2×2 weighted average. The default for continuous imagery (DEMs, RGB photos).
• ResamplingMethod::Cubic — 4×4 cubic-convolution kernel. Sharper than bilinear at moderate extra cost.
• ResamplingMethod::Lanczos — 6×6 windowed-sinc. Highest quality; slowest. Used for final-output downsampling where artefacts matter.
• ResamplingMethod::Average — area-weighted mean. Best for building pyramid overviews of continuous data (preserves means under downsampling, unlike bilinear).
• ResamplingMethod::Mode — most-frequent value within target area. For categorical pyramid overviews.

Functions

resample_tile

Resize a RasterTile to new (target_width, target_height). Used by pyramid building and warp output preparation.

sample

Sample one fractional (x, y) from band using the chosen method.