gradient operators

Terms from Artificial Intelligence: humans at the heart of algorithms

Page numbers are for draft copy at present; they will be replaced with correct numbers when final book is formatted. Chapter numbers are correct and will not change now.

A gradient operators or gradient filters operates on an image to measure the gradient, that is the rate and direction of change of an image at each point. They are often used for edge detectoon early in a computer vision pipeline. For example, areas that change form dark to light have a gradient orthogonal to the line between the dark and light areas. Some are simple linear filters that produce separate values for the rates of change in x and y dirctions separately (the gradient vector). However, for initial edge detection we sometimes just want to measure the overall size of the gradient; gradient operators, such as Robert's operator and Sobel's operator focus on this.

Defined on page 253

Used on Chap. 12: pages 253, 256, 276

Also known as gradient filters

Different gradient filters: (i) and (ii) simple x-dircetion filters; (iii) and (iv) x and y versions of a slightly more complex filter.

Applying a gradient filter

Gradient filter on a diagonal edge