Overview: Scaling kernels

Scaling kernels define how neighboring source samples contribute to a reconstructed output sample.

Nearest-neighbor, linear and higher-order filters assign different weights and neighborhoods to the reconstruction. Those choices influence sharpness, smoothing, ringing and aliasing. The individual kernel pages show the weighting and its visible consequences, making them useful comparisons for an actual scaling task.

• Nearest-neighbor scaling
  Nearest-neighbor scaling assigns each output position the value of the closest source sample.

• Bilinear scaling
  Bilinear scaling interpolates an output value from the four neighboring source positions using linear weights on each axis.

• Bicubic scaling
  Bicubic scaling uses cubic weight functions along two axes to reconstruct output values from a wider source neighborhood.

• Lanczos scaling
  Lanczos scaling reconstructs samples with a finite windowed-sinc kernel whose radius controls the contributing neighborhood.

• EWA Lanczos (Jinc)
  EWA Lanczos uses a radial Jinc-based reconstruction footprint to combine source samples around an output position.

• Spline scaling
  Spline scaling reconstructs picture values with piecewise polynomial weight functions whose support and shape depend on the specified spline family.