pipeline

The Image Processing Pipeline

An image processing pipeline is the representation of an ordered list of operations, and is a complete description of the processed result for any particular original image.

Pipelines are identified by the parameters that describe them, for example, a parameter list saturation=10,width=500 will correspond to the following pipeline:

Pipeline {
	Resize {
		Width: 500,
	},
	Adjust {
		Saturation: 10,
	},
}

The above is a complete description of the manipulations required for processing an image. Also, note that the order of parameters does not affect the final order of operations, and thus a pipeline can be identified by more than one parameter lists.

Parameters are comma-separated key-value assignments, for example width=500,fit=crop. Certain parameters have additional constraints on their values, as described below.

Operations

Operations are the building blocks of the image processing pipeline, and are defined as sets of related image manipulation tasks, e.g. resizing, adjusting colors etc.

What follows is a reference list of all available operations, along with a list of parameters relevant to each one.

Resize

The resize operation handles any manipulation of the image's dimensions, including clipping and cropping. The parameters relevant to this operation are:

width and height

These parameters accept any integer value, but negative numbers and values that are equal or exceed the original image's resolution result in the original image being returned.

fit

Determines the way in which the image will attempt fit the constraints imposed by the pipeline. Supported fit modes and their additional options include:

• clip: Attempts to resize image so that resulting image dimensions are smaller or equal to the pipeline constraints. So, for an image of size 1000x500 and a pipeline of width=500,height=200, the resulting image will be of size 400x200. This is the default.
• crop: Attempts resize image to the exact size requested, cropping any additional parts of the image. Supports the following colon-separated options:
  • top, bottom, left, right, center, which define the center of gravity for the cropped image. So, for the above example and a fit of fit=crop:bottom, the top 50 pixels of the image would be cropped. Default is center.
  • point, which defines the center of gravity for a cropped image as X and Y pixel co-ordinates. For example, the center point of focus for the above example would be expressed by a pipeline of fit=crop:point:500:250.