Low-level feature extraction (including edge detection) - Feature Extraction and Image Processing - page 100

Table 4.1 Overview of Chapter 4

Main topic Sub topics Main points

First-order What is an edge and how we detect Difference operation; Roberts

edge it. The equivalence of operators to Cross, Smoothing, Prewitt, Sobel,

detection first-order differentiation and the Canny.

insight this brings. The need for
filtering and more sophisticated
first-order operators.

Second- Relationship between first- and Second-order differencing;

order edge second-order differencing operations. Laplacian, Zero-crossing

detection The basis of a second-order operator. detection; Marr-Hildreth,

The need to include filtering and Laplacian of Gaussian.

better operations.

Other edge Alternative approaches and perfor- Other noise models: Spacek.

operators mance aspects. Comparing different Other edge models; Petrou.

operators.

Detecting Nature of curvature . Planar curvature; corners.

image Computing curvature from: edge Curvature estimation by: change

curvature information; by using curve approxi- in edge direction; curve fitting;

mation ; by change in intensity; and intensity change; Harris corner

by correlation . detector.

Optical Movement and the nature of optical Detection by differencing. Optical

flow flow. Estimating the optical flow by flow; aperture problem;

estimation differential approach. Need for smoothness constraint.

other approaches (including Differential approach; Horn and

matching regions). Schunk method; correlation.

detection is insensitive to change in the overall illumination level. Edge detection. highlights
image contrast . Detecting contrast, which is difference in intensity, can emphasise the
boundaries of features within an image, since this is where image contrast occurs. This is,
naturally, how human vision can perceive the perimeter of an object, since the object is of
different intensity to its surroundings. Essentially, the boundary of an object is a step-
change in the intensity levels. The edge is at the position of the step-change. To detect the
edge position we can use first-order differentiation since this emphasises change; first-
order differentiation gives no response when applied to signals that do not change. The
first edge detection operators to be studied here are group operators which aim to deliver
an output which approximates the result of first-order differentiation.
A change in intensity can be revealed by differencing adjacent points. Differencing
horizontally adjacent points will detect vertical changes in intensity and is often called a
horizontal edge detector by virtue of its action. A horizontal operator will not show up
horizontal changes in intensity since the difference is zero. When applied to an image P the
action of the horizontal edge detector forms the difference between two horizontally adjacent
points, as such detecting the vertical edges, Ex , as:
Ex x , y = | P x , y - P x +1, y | ∀ x ∈ 1, N - 1; y ∈ 1, N (4.1)
In order to detect horizontal edges we need a vertical edge detector which differences