Module gradient 

Gradient operators for QSM

Forward difference gradient and backward divergence operators used in TV regularization and other algorithms.

Functions

bdiv

Backward divergence operator (negative adjoint of forward gradient)

bdiv_inplace

Backward divergence operator (in-place)

bdiv_inplace_f32

Backward divergence operator (in-place, f32) Uses zero boundary conditions (matching Julia)

bdiv_masked_inplace_f32

Backward divergence operator (in-place, f32) - masked version

fgrad

Forward difference gradient operator

fgrad_f32

Forward difference gradient operator (allocating, f32)

fgrad_inplace

Forward difference gradient operator (in-place)

fgrad_inplace_f32

Forward difference gradient operator (in-place, f32)

fgrad_masked_inplace_f32

Forward difference gradient operator (in-place, f32) - masked version Only computes gradient where mask is non-zero

grad_magnitude_squared

Compute gradient magnitude squared: |∇x|² = gx² + gy² + gz²

symdiv_inplace_f32

Divergence of symmetric tensor field (adjoint of symgrad)

symgrad_inplace_f32

Symmetric gradient operator for TGV regularization (in-place, f32)