Interactive registration

It is possible to manually transform a neuroglancer layer using ngtools.

Interactive shifts

The active layer can be transformed using predefined key presses. We define the following shortcuts:

As soon as one of these actions is triggered, three sliders are defined in the shader of the active layer. These sliders allow the scale of each type of action (translation, rotation or zoom) to be tuned. These sliders are defined via

#uicontrol float translation_scale slider(min=-5, max=5, default=0, step=0.01)
#uicontrol float rotation_scale slider(min=0, max=90, default=1, step=0.1)
#uicontrol float zoom_scale slider(min=-5, max=5, default=0, step=0.01)

and look like this:

• The rotation slider defines the number of degrees applied by one key press.
• Translation are expressed in terms of "model space units", which are usually defined by the output dimensions of the topmost layer. These units are scaled by 2 ** translation_scale. Therefore, when translation_scale == 0, each key press applies a translation of one model unit; when translation_scale == 1, each key press applies a translation of two model units; and when translation_scale == -1, each key press applies a translation half a model unit.
• The zoom slider defines the numbers of powers of two (or one over two) applied by each zoom or unzzom key press. When zoom_scale == 0, each key press zooms by a factor 2 (or unzoom by a factor 0.5); when zoom_scale == 1, each key press zooms by a factor 4; and when zoom_scale == -1, each key press zooms by a factor sqrt(2).

Cursor-based translation

The shortcut Ctrl+T applies a translation that brings the location pointed by the mouse cursor to the location of the crosshair. This action is named translate-layer-mouse-to-crosshair.

Landmark-based registration

It is also possible to apply a translation, rigid body transformation, similtude or affine transform that minimizes the distance between two (paired) point clouds. The point clouds can be defined (or reset) using the following shortcuts:

These actions automatically create two annotation layers named ::landmarks::fixed and ::landmarks::moving. These layers can be deleted after use, or if one wishes to clear existing landmarks.

Once the two pointclouds are defined, a transformation can be fitted and applied using any of these shortcuts:

The moving point cloud also gets transformed, and can then be augmented with additional landmarks and re-used.