MAPaint Technical Report:
Painting

Painting

The basic function of the paint program is to allow interactive segmentation of a reference image to generate a series if binary images which map directly onto the original reference and (presumably) correspond to structures of some sort within the original. At the lowest level the segmentation is by interactively marking - painting - the reference image to delineate the region of interest. This is not actually changing the original, merely creating a binary mask which is displayed a coloured ``wash'' over the original grey-values. Within the paint program it is possible to delineate up to five different regions at the same time, more than that is difficult because the program is designed to run with the colours displayed using overlay planes for efficiency. On an 8-bit colour display with a reduced colour availability and the requirement that the overlays appear as a wash over the grey-values this effectively reduces the number of possible simultaneous paint domains to five - three overlay planes allows 32 colours per wash, reserve 32 colours for other applications, reserve 32 values for the grey-values leaving 5x32 for the overlays plus 32 colours for ``solid'' overlays. In the current version the user can select upt to 5 ``wash'' overlays and 27 ``solid overlays.

For the atlas it will often be the case that the edge of one structure will delineate the boundary of the adjacent structure and therefore for efficiency and consistency each boundary should be defined just once. To aid this process the program will enforce boolean ``rules'' on the domains defined interactively as a simple dominance hierarchy. By default the overlays are ordered so that domain-1 has dominance over domain-2, domain-2 has dominance over domain-3 and so on. Within the program there is the option to change this dominance ordering (by ``drag and drop'') so that the user can modify the behavior as required.

There are currently two painting modes. The first is referred to as ``drawing'' in which the user draws a boundary, either by dragging which will result in an arbitrary polyline or by point selection which will result in a line consisting of straight-line segments. Once the polyline is complete the inside of the contour defines a region that is used to modify the current domain. The new domain will be added (set union) or subtracted (set difference) according depending on which button is pressed or if the modifier key Meta is pressed.

The second mode is to use the ``paint-ball'' which is equivalent to a ``paint-brush'' and ``eraser'' depending on which button is pressed or if the meta modifier key is pressed. With both of these options there are a variety of parameters - cursor appearence, paintball size and shape. Painting a screen resolution is very demanding therefore is is possible to temporarily bring up magnification windows to any level which allows painting to single pixel level when necessary. In the magnified windows the painted domain is constrained to the pixel regions of the base image and will therefore have a ``blocky'' appearence.

The default parameters controlling the painting and drawing process can be controlled from a tool control dialog. This dialog switches as the selected painting tool changes to reflect the parameters available for that option. In the case of drawing th only options area choice oc cursor - single pixel dot, cross bars or arrow. Usage to date suggests that the single dot is most useful and is therefore set as the default.

The supplementary parameters with the drawing option are paint-ball size, paint-ball shape (circle or square) and a choice of contrasting boundary for the paint-ball which can be useful on dark images.