Application of an improved method of Wavelet Transform in image edge detection
Shulei Wu
1
, Zongan Zhang
1
, Huandong Chen
1,*
1
College of Information Science and Technology
Hainan Normal University
HaiKou, China
E-mail: wus hulei@neteas e.com
Jinmei Zhan
2
2
College of Information Science and Technology
Hainan University
HaiKou, China
* Corresponding author e-mail:
ch_huandong@163.com
Abstract— Edge detection is widely used to extract and analyze
contour and texture feature. This paper focuses on researches
of wavelet analysis and the application of its improved method
in edge detection. The basic idea of the improved method is
that we first preprocess image, and the image texture will be
shown when processing it under a uniform threshold, then the
texture curve becomes thinner after the morphological
processing for the texture is done. Experimental results show
that the improved texture is more clearly and accurately under
the same threshold and image texture becomes much thinner
after morphological preprocessing. The improved method is
effective.
Keywords-Edge detection; Wavelet; Texture
I. INTRODUCT ION
Edge detection and spectral analysis are bases of
computer vision and information processing. With the
progress of science technology, many algorithms have been
proposed, which is classified into two kinds: one is classical
edge operator, the other is based on mathematical
morphology[1-2] and edge fuzzy theory. The classical edge
operator, based on spatial operation, gets edge information
by convolution of original image pixels and operators, such
as Roberts and Canny operator etc. The mathematical
morphology based edge detection method [3] extracts image
shape to analyze and identify the image. The fuzzy theory
based edge detection method achieved good results in image
proces s ing. But it s till has much s pace to improve.
Wavelet is a multi-scale transform, which can extract
useful information effectively. Compared with the traditional
operator, wavelet not only retains important details, but also
removes a large amount of false information, which develops
an effective tool for the edge detection and spectrum analysis.
However, wavelet also has obvious shortcomings, for
example, weak texture region cannot be displayed, and that
some non-texture curve will appear when a bigger threshold
is given. In addition, the edge of wavelet is wider and the
texture is heavy. Thus, it is very important to improve
wavelet.
II.
PRINCIPLE OF IMAGE EDGE DET ECTION BASED ON
WAVELET TRANSFORM
Wavelet transform decomposes a signal[4] into several
parts of multi-dimension. Choose appropriate wavelet
function to make mutation points of the edge more actual.
Wavelet transform based edge detection[5] uses
smoothing function
˄W˅
. Smooth detected signal under
each scale obtains mutation points by once or the second
differential. The inflection point of a smooth signal is a
maximum of a differential or zero crossing point of the
second differential[6]. Suppose wavelet function
is
G˄W˅
˄W˅
GW
, the edge detection is achieved by the
extreme value of function coefficient.
Suppose two-dimensional image signal is
[
I
and two-
dimensional smoothing function is
[\
, which satisfies
formula (1) (2):
[\G[G\
OLP [\
[
Perform s moothing processing for the original image,
after processing:
[\ [\ I[\
J
Calculate the first order differential for smoothed image:
[\
[\ I[\
J
[[
[\
[\
I
[
[\
[\ I[\
J
\
[\
[\
I
[
2015 11th International Conference on Computational Intelligence and Security
978-1-4673-8660-9/15 $31.00 © 2015 IEEE
DOI 10.1109/CIS.2015.114
450
2015 11th International Conference on Computational Intelligence and Security
978-1-4673-8660-9/15 $31.00 © 2015 IEEE
DOI 10.1109/CIS.2015.114
450
2015 11th International Conference on Computational Intelligence and Security
978-1-4673-8660-9/15 $31.00 © 2015 IEEE
DOI 10.1109/CIS.2015.114
450
2015 11th International Conference on Computational Intelligence and Security
978-1-4673-8660-9/15 $31.00 © 2015 IEEE
DOI 10.1109/CIS.2015.114
450
2015 11th International Conference on Computational Intelligence and Security
978-1-4673-8660-9/15 $31.00 © 2015 IEEE
DOI 10.1109/CIS.2015.114
450
2015 11th International Conference on Computational Intelligence and Security
978-1-4673-8660-9/15 $31.00 © 2015 IEEE
DOI 10.1109/CIS.2015.114
450