Module: Registration 3D-3D ()

Description:

Registers a pair of 3D tomograms.

Connections:

Transformed Data [required]
The data to transform by registration. Limitation: only 16-bit unsigned data with less than 2 billion voxels are currently supported.

Fixed Data [required]
The reference data for performing the registration. Limitation: only 16-bit unsigned data with less than 2 billion voxels are currently supported.

Ports:

Transformation Type

Type of transformation to use when searching for optimal alignment: SIMILARITY: A 7 parameter/degrees-of-freedom transform: translation (3 DOF), rotation (3 DOF) and isotropic scaling (1 DOF). AFFINE: A 12 parameter/degrees-of-freedom transform: translation (3 DOF), rotation (3 DOF), anisotropic scaling (3 DOF) and shear (3 DOF).

Registration Task

Indicates the types of data which are to be registered, used as a guide for selecting parameters. 3D with 3D: Register pair of tomograms. This task allows selection from the full range of parameters. wet 3D with dry 3D: Register pair of tomograms, one flooded with contrast, other dry. Input data should be the saturated/flooded/wet tomogram, fixed image data should be dry tomogram. high 3D with low 3D: Register different resolution tomograms, for example input data is tomogram of 8mm core, fixed-image data is tomogram of 38mm core.

Registration Metric

Register pair of tomograms. This task allows selection from the full range of parameters.

Registration Metric

Register pair of tomograms, one flooded with contrast, other dry. Input data should be the saturated/flooded/wet tomogram, fixed image data should be dry tomogram..

Registration Metric

Register different resolution tomograms, for example input data is tomogram of 8mm core, fixed-image data is tomogram of 38mm core. .

Max Down Sample Factor

Set the downsampling factor used to register images at lowest resolution. Generally, a factor of 16 is good for 2048x2048x2048 tomographic images.

Optimizer Tolerance

Stop tolerance for the cost function value, when relative difference of cost function is less than this tolerance the optimization iterations are stopped. Typically the registration transformation parameters are at within sqrt(optimizer_tolerance) of the global minimum. The larger the value of this tolerance, the fewer iterations required to achieve convergence, i.e. faster runtime.

Dense Sampling Refinement

If true, performs an extra refinement registration at full-resolution using the 'dense' point-sampling specified in the point_sample parameter. If false, skips the refinement registration. Performing the full-resolution dense refinement registration is often time consuming and is usually unnecessary as it results in negligable changes to the registration parameters. One can perform the refinement in a subsequent chained operation using Exhaustive Initial Search and Dense Sampling Refinement.

Minimum Overlap Percent

This is a helper parameter used to indicate the volume of overlap of the image pair in the aligned state. It is used in two ways:

• to determine the range of translations in the initial basic search, the higher the Minimum Overlap Percent the smaller the translation range and hence a faster runtime for the initial-exhaustive-search.
• to avoid spurious registrations, where a small percentage of overlapping voxels can generate a good/low cost-function value (e.g. a high correlation coefficient). The Minimum Overlap Percent is the threshold-percentage of overlapping "voxels" at which a penalty term is added to the cost function and this penalty term discourages selection of alignments/registrations where there is little overlap of non-masked volume. If the percentage of overlapping voxels is greater than Minimum Overlap Percent, then a penalty term is not added. The penalty will only be used/effective if the fixed image is masked.

Overlap percent is calculated using the number of non-masked voxels in the transform-image which overlap with non-masked voxels in the fixed-image (divided by the total number of non-masked voxels in the transform-image).

In cases where the volume/number-of-non-masked-voxels in the transform-image exceeds that of the fixed-image (because the transform-image contains a greater portion of the imaged specimen) then the Minimum Overlap Percent should be adjusted accordingly. If the specimen is a cylinder of radius and height , and the fixed-image contains a portion of the cylinder which is with the transform-image containing the full length of the specimen, then the Minimum Overlap Percent should be set less than (or equal to) 50

Try Flipped Orientation

This parameter is used when auto-calculating the initial basic/exhaustive search rotations. If true, a 180-degree rotation about the x-axis is included in the possible orientations of the initial-basic-search. Required if one of the images was acquired "upside down" relative to the other. A value of true will double the runtime of the initial exhaustive search (compared with a value of false).

Tilt Angle Degrees

The search range for rotation in the "tilt" direction (rotation about the non-z-axes). The range is in the interval Increases runtime by a factor of

Exhaustive Initial Search

If true, performs initial exhaustive lowest resolution registration-search of the input transform-tomogram with the fixed-tomogram. If false, the exhaustive search is skipped. Skipping the initial exhaustive search registration is useful for refining an existing/approximate registration.

Transform Image Mask

Defines the intensities which are masked in the transform image. Wise to at least do a shrink-wrap-mask/geometric-mask on the "fixed image" and use the mask value option.

Transform Image Mask Value

This single intensity value is used to mask voxels

Transform Image Mask Interval

Voxels with intensity values inside the range are masked.

Transform Image Mask Non Interval Min

Voxels with intensity values less than or equal to the value are masked.

Transform Image Mask Non Interval Max

Voxels with intensity values greater than or equal to the value are masked.

Transform Image Mask

Defines the intensities which are masked in the fixed image. Wise to at least do a shrink-wrap-mask/geometric-mask on the "fixed image" and use the mask value option.

Transform Fixed Image Mask Value

This single intensity value is used to mask voxels

Transform Fixed Image Mask Interval

Voxels with intensity values inside the range are masked.

Transform Fixed Image Mask Non Interval Min

Voxels with intensity values less than or equal to the value are masked.

Transform Fixed Image Mask Non Interval Max

Voxels with intensity values greater than or equal to the value are masked.

FFT Correlation Initial Search

Perform FFT correlation for the initial exhaustive search. Provide significant speedup for discovering approximate alignments.

FFT Correlation

Use phase-correlation which favours matching of edges within the images.