dic_preprocess.analysisRun.alignMovingToReference

Inputs

referenceImage

Reference image whose height and width define the output canvas.

movingImage

Numeric grayscale or multichannel image to transform.

fixedPoints

N-by-2 [x y] coordinates in referenceImage.

movingPoints

N-by-2 corresponding [x y] coordinates in movingImage. Use the same number and order of points as fixedPoints. At least two distinct pairs are needed to constrain a 2-D rigid transform.

Outputs

alignedImage

movingImage resampled onto the reference canvas. It has the moving image class and channel count; pixels outside the source are 0.

tformRigid

Three-by-three row-vector homogeneous transform, shown as transform in the usage syntax. A moving coordinate maps to reference coordinates as [x y 1]*tformRigid.

Description

The function uses the least-squares orthogonal Procrustes solution for one rotation and translation; reflection is explicitly rejected. It does not estimate scale or shear. Resampling is delegated to applyRigidTransform.

Failure Behavior

The caller is responsible for supplying corresponding finite N-by-2 point arrays with enough nondegenerate geometry to constrain the fit. The function does not validate pair count or fit quality; insufficient or degenerate data can therefore return an arbitrary transform. Shapes that cannot participate in the matrix operations, unsupported image classes, and SVD/interpolation failures propagate the originating MATLAB error.

Example

reference = zeros(12, 12);
moving = zeros(12, 12); moving(4:6, 3:5) = 1;
movingPoints = [3 4; 5 4; 3 6];
fixedPoints = movingPoints + [2 1];
[aligned, transform] = ...
dic_preprocess.analysisRun.alignMovingToReference( ...
reference, moving, fixedPoints, movingPoints);
assert(isequal(size(aligned), size(reference)))
assert(norm([movingPoints ones(3,1)]*transform - ...
[fixedPoints ones(3,1)], "fro") < 1e-10)