kernelkit.kernels.RayDrivenConeFP

class kernelkit.kernels.RayDrivenConeFP(pixels_per_block: tuple = None, slices_per_block: int = None, volume_axes: tuple = (0, 1, 2), projection_axes: tuple = (0, 1, 2))

Ray-driven conebeam forward projection kernel.

Attributes:

cuda_source

Returns the CUDA source code of the kernel

is_compiled

projection_axes

volume_axes

Methods

FLOAT_DTYPE	alias of float32
__call__(volume_texture, ...[, ...])	Forward projection with conebeam geometry.
geoms2params(projection_geometry, ...)	Converts geometries to kernel parameters.
set_params(params)	Transfer geometries to device as structure of arrays.

compile

__init__(pixels_per_block: tuple = None, slices_per_block: int = None, volume_axes: tuple = (0, 1, 2), projection_axes: tuple = (0, 1, 2))

Conebeam forward projection kernel.

Parameters:

pixels_per_blocktuple, optional

Number of pixels per block, defined as (angles, rows, columns).

slices_per_blockint, optional

Number of slices per block in the direction of the ray. Defaults to 16.

volume_axestuple, optional

Axes of the reconstruction volume. Defaults to (0, 1, 2). The axes correspond to ‘x’, ‘y’, and ‘z’, respectively. ‘x’ is the source-detector axis, ‘y’ the orthogonal axis in the horizontal plane, and ‘z’ the vertical axis. ASTRA Toolbox uses (2, 1, 0).

projection_axestuple, optional

Axes of the projection data. Defaults to (0, 1, 2). Axis 0 corresponds to the projection angle, axis 1 to the detector column, and axis 2 to the detector row. ASTRA Toolbox uses (1, 0, 2) for this parameter.

*args

Arguments passed to the Kernel constructor.

See also

kernelkit.kernel.BaseKernel

Base class for CUDA kernels.

Methods

__init__([pixels_per_block, ...])	Conebeam forward projection kernel.
compile([max_projs])
geoms2params(projection_geometry, ...)	Converts geometries to kernel parameters.
set_params(params)	Transfer geometries to device as structure of arrays.

Attributes

SUPPORTED_DTYPES
cuda_source	Returns the CUDA source code of the kernel
is_compiled
pixels_per_block
projection_axes
slices_per_block
volume_axes