Please enable JavaScript.
Coggle requires JavaScript to display documents.
Alignment method (Method description (input volumetric data is loaded into…
Alignment method
input volumetric data is loaded into memory; the user determines the sequence of rotational planes by the name of orthogonal axis to that planes
data is filtered with 2D gaussian of size of smallest features; data is segmented in slice-wise fashion with combinations thresholding methods; checking over-segmentation at each slice; sequence of morphological operations are applied (several iterations of closing and filling holes)
- when segmented;
- for each rotation plane, we obtain a projection of the intensities of segmented data on this plane;
- to do that, perform a logic OR operation of values in a row obtained along projection direction, that guarantees if the row has at least one non-zero value, the result will give you non-zero value as well;
-
then, the produced eigenvectors should be rotated to be aligned with reference orthogonal axes defined at the same plane
after that, the original dataset is rotated by the determined angle in the corresponding plane
now, the process is repeated, but over the changed dataset
when the dataset will be processed from all projection planes, it will be aligned with the defined orthogonal axes which we took as a reference.
The proposed method allows us to align large datasets without placing completely them into the memory. Instead sequentially aligning them by rotation in the corresponding projection planes, in this way, the processing occurs slice-by-slice which shortcut memory and computational burdens.
- [Intro Description] The proposed method is based on
-
-
-
-
-
when there is no template data, registration cannot be applied
-
-