18342075-米家龙-计算机视觉期中作业-Edge detection1
需积分: 0 35 浏览量
2022-08-03
12:29:58
上传
评论
收藏 1.56MB PDF 举报
Lecture Node for Edge Detection
mijialong
Abstract
This learning report is used to record my learning in
edge detection. In this report, I will record some common
algorithms used in edge dection. For example, The Sobel
operator, sometimes called the Sobel–Feldman operator or
Sobel filter, Scharr operator, Canny edge detector and
Deriche edge detector.
In addition to the algorithm principle and specific steps,
I will also use the code corresponding to each algorithm
to process some test image to see the effect of each edge
detection algorithm.
1. Introduction
Now computer vision developing rapidly, image process-
ing technology has been applied to all aspects of our lives.
As one of the more basic directions in computer vision,
edge detection has a history of decades of development.
1.1. Edge detection
Edge detection is a basic problem in computer vision
and image processing, which is aim at identifying points
that brightness in image changes sharply, or more formally.
This problem includes a varity of mathematical methods
and most of methods are suitable for using programming
language to write code and hand it over to the computer for
implementation.
1.2. Edge
The sharp changes between adjacent points can be used
to shown important events and changes. These changes of
image brightness are likely to correspond to:[1]
• discontinuities in depth
• discontinuities in surface orientation
• changes in material properties
• variations in scene illumination
The edge can be classified as viewpoint-dependent or
viewpoint-independent, which extracted from a 2D image
of a 3D scene. A viewpoint-dependent edge may change
with the different viewpoints and offen shows the grometry
of the scene. AAnd the properties of 3D objects usually can
be reflected by a viewpoint-independent edge.
1.3. Effect
Applying edge detection to a image process may result
a set of connected curves that the indicate the boundaries
so that can typically reduce the amount of the data in the
image that need to be processed. And it can filter out much
more information, which is considered as less as relevant,
while save the structural charateristics that is important of
the image.
Edge detection has been applied into image processing,
computer vision and machine vision in the academic field,
which is in the fileds of feature detection and feature extrac-
tion. It is also a basic step in image analeysis, image pattern
recognition and image processing.
2. Edge detector
There are lots of methods for edge detection, and most of
them can be classified into 2 catagories: a) Based on search;
b) based on zero crossing.
The search based method usually calculates the measures
of edge strength with a first-order derivative expression such
as gradient magnitude firstly, and then uses the calculated
estimate of the local orientation of the edge, usually the gra-
dient direction, to search for the local directional maximum
of the gradient magnitude. [3]
The zero-crossing based methods search for zero-
crossing in a second-order derivative expression to find
edges which calculated from the image. The common zero-
crossing expressions are the zero-crossing of the Laplacian
operator or the zero-crossing of the nonlinear differential
expression.
A smoothing stage is almost applied as a pre-processing
step for edge detection, which is usually Gaussian smooth-
ing, to reduce image noise.
2.1. Image gradient
Define the gradient of an image:
∇f = [
∂f
∂x
,
∂f
∂y
]
1
评论0