An image can either be a file, an image loaded into an underlying library or a
simple buffer of pixels. Each of these states has its own Python class
RGBAImageBuffer are three
of the image classes in Willow.
These are functions that perform actions on an image in a particular state. For
Operations can either be defined as methods on the image class or as functions registered separately.
All operations are registered in a central registry and will appear as a method on all other image classes. If it’s called from a class that doesn’t implement the operation, the image will be automatically converted to the nearest image class that supports it and the operation is run on that.
Operations that alter an image return a new image object instead of altering the source one. This also means that if a conversion took place, the new image’s class would be different.
These are functions that convert an image between two image classes. For
example, a converter from
PillowImage would simply be a
function that calls
PIL.Image.open on the underlying file to get a Pillow
Like operations, these can either be methods on the image class or registered separately.
Each converter has a cost which helps Willow decide which is the best available image library to use for a particular file format.
The registry is where all image classes, operations and converters are registered. It contains methods to allow you to register new items and even override existing ones.
It also is responsible for finding operations and planning routes between image classes.
These are used to group related image classes, operations and converters together allowing them to be registered as a single unit.
The convention within Willow is to create a single plugin for each underlying library. The default ones are “pillow”, “wand” and “opencv”.
Plugins can be registered even if the underlying library is not installed. This allows Willow to generate a useful error message if an operation is requested that only exists in a plugin without an underlying library.