Hello,

It usually gives better results having the alignment done first, and then running edge-extraction on the colorful atlas slice at the end. The rationale behind this is if the outlines are already present at the beginning, as part of the atlas itself, when adjusting the atlas slice to fit the actual image, the pixels of the outline will be stretched-distorted too. While creating the outline as last step allows having a fixed, uniform thickness for the outline.

Image registration itself depends a lot on the image, for example if it's an actual 3D volume (MRI, PET, and the like), a series of images that can be reconstructed to a 3D volume (e.g. block-face methods), or 2D section images which need individual adjustments (e.g. histology without block-face support).

Best regards,

Gergely