This suggests that in the simplest way one could calculate curvature by computing the first and second derivatives of the x- and y - components of the surface.
The surface computation of curvature involves fitting a quadratic surface to the mapped horizon using a least-squares regression and using nine sample points (eight neighbors around a given point). When nine sample points are used for the computation of curvature, it results in an overdetermined system when the solution is sought by the least squares regression method. Different measures of curvature can then be written in terms of the six coefficients, as was shown by Roberts (2001).
For
Of all the available curvature measures, Chopra and Marfurt (2007) recommend the application of the most-positive curvature and most-negative curvature attributes, for they are the easiest to understand intuitively.
Roberts (2001) has given the following expression for computation of the most-positive curvature.
where a, b and c are the coefficients of the polynomial equation.
As we note above, 
in the Fourier domain is equivalent to multiplying the spectrum by (
Both such estimates have their applications, long wavelength suitable for obtaining the gross definitions of the geometrical features and the short wavelength for extracting the finer details and thus more resolved images.
Notice the clear set of faults/fractures generated from coherence and curvature attributes