dic_preprocess.analysisRun.autoAlignMovingToReference

Inputs

referenceImage

Numeric grayscale or RGB reference image. Its first two dimensions define the output canvas and correlation size.

movingImage

Numeric grayscale or RGB image to translate.

Outputs

alignedImage

Original movingImage translated onto the reference canvas, with linear interpolation and zero fill.

tformRigid

Three-by-three row-vector homogeneous translation transform, shown as transform in the usage syntax.

method

Character vector identifying the fixed phase-correlation method.

Description

Each image is converted to normalized grayscale independently. For shift estimation only, moving grayscale data is resized to the reference size by nearest-neighbor sampling. Phase correlation returns a whole-pixel circular shift. Rotation, scale, deformation, repeated texture, and large nonoverlap can produce a poor fit.

Failure Behavior

The function does not assign a confidence score or reject an ambiguous phase-correlation peak; low-texture or repeated-pattern inputs can return a numerically valid but poor translation. Empty arrays, unsupported image classes, or invalid channel shapes propagate image conversion/interpolation errors.

Example

reference = zeros(16); reference(5:8, 6:9) = 1;
moving = circshift(reference, [2 -3]);
[aligned, transform, method] = ...
dic_preprocess.analysisRun.autoAlignMovingToReference( ...
reference, moving);
assert(isequal(size(aligned), size(reference)))
assert(isequal(size(transform), [3 3]) && contains(method, "phase-correlation"))