The Sentinel-2 level-1C raster data in cartographic geometry is resampled from level-1B data in sensor geometry, using B-spline interpolation. Which order of B-spline interpolation is used?

ESA Sentinel Online states just that:

Radiometric Interpolation

Interpolation estimates the radiance values of the target point, knowing the radiance of neighbouring pixels. A linear algorithm is used for this operation with B-spline functions.

and I haven't been able to find the information by googling.

1 Answer 1


Based on this quote from Javier Gorroño; Andrew Banks; Ferran Gascon; Nigel P. Fox; Craig I. Underwood Novel techniques for the analysis of the TOA radiometric uncertainty Proc. SPIE 10000, Sensors, Systems, and Next-Generation Satellites XX, 100001H (19 October 2016):

The last method are the B-splines interpolation that is implemented as the nominal resampling method for the S2 L1C products17, 21. Both cubic convolution and B-splines require a 4×4 kernel grid that produces a smoother image at the expense of a more demanding computational requirement20, 21.

it's a cubic B-spline, based on that the support of the impulse response of cubic B-spline is of length 4.

In one dimension the impulse response of cubic B-spline is (grid lines at integer values):

enter image description here

Cubic B-spline interpolation can be implemented by the following pseudocode with input x, position p = i + f, integer position i, fractional position 0 <= f <= 1 and output y:

c0 = 1/6.0*(x[i - 1] + x[i + 1]) + 2/3.0*x[i]
c1 = 1/2.0*(x[i + 1] - x[i - 1])
c2 = 1/2.0*(x[i - 1] + x[i + 1]) - x[i]
c3 = 1/2.0*(x[i] - x[i + 1]) + 1/6.0*(x[i + 2] - x[i - 1])
y = ((c3*f + c2)*f + c1)*f + c0

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.