To obtain the best output quality tiles from the MapTiler Engine application, it is possible to choose a resampling method to be used in the rendering process. In this article, you will learn what resampling is, how the selection of the resampling method affects the output, and where this setting can be made from.
The global settings dialog is available since MapTiler Engine 11.3.
Resampling is the process of interpolating pixel values while transforming your raster dataset. This is used when the input and output do not line up exactly, when the pixel size changes, when the data is shifted, or a combination of these.
The selected resampling method affects the visual quality of the output tiles. As mentioned above, it is used for interpolation of the values of individual pixels and it affects the sharpness vs smoothness of the produced maps.
Resampling methods in MapTiler Engine
In MapTiler Engine there are 6 resampling methods that you can use in the rendering process.
You can find them in MapTiler Engine / Preferences / Global Settings / Image / Resampling.
Available methods for resampling
The following section describes each of the available resampling methods and corresponding output using the same input.
Performs the nearest neighbor assignment and is the fastest of the interpolation methods. It is used primarily for discrete data, such as land-use classification. Rarely makes sense for production data can be used for quick testing since it is much faster than the others.
This is the default option for resampling. Performs a bilinear interpolation and determines the new value of a cell-based on a weighted distance average of the four nearest input cell centers (2x2 pixel kernel). It is useful for continuous data and will cause some smoothing of the data.
Performs a cubic convolution and determines the new value of a cell-based on fitting a smooth curve through the 16 nearest input cell centers (4x4 pixel kernel). It is appropriate for continuous data, although it may result in the output raster containing values outside the range of the input raster. If this is unacceptable, use Bilinear instead. The output from cubic convolution is geometrically less distorted than the raster achieved by running the nearest neighbor resampling algorithm. The disadvantage of the Cubic option is that it requires more processing time.
Cubic B-splines use an interpolation kernel that is the result of convolving a unit-width box with itself 4 times. This has some nice properties, but simple B-spline interpolation gives an output function that does not pass through the original sample points. Each output point depends on 16 input points (4x4 pixel kernel), as before.
Average resampling computes the average of all non-NODATA contributing pixels.
Mode resampling selects the value which appears most often of all the sampled points.
Available methods for resampling overviews
Mostly used for elevation maps or similar.
This is the default option for overviews resampling, which computes the average of all non-NODATA contributing pixels.
A lifetime of the setting
Until the first time you set the other than the default value, the default values are used. After that, selected options are saved and loaded on the next application startup. Therefore if you would like to return to the default resampling values after you once performed the selection, you would need to click "Reset defaults" in the bottom panel of the Global settings dialog.
This article explained the influence of the used resampling method on the output and showed where this setting is located in the application.