cient addi-
ssifications
let and the
isual qual-
sults based
K
brightness
pectral sig-
1. In order
ose to per-
atic image,
n a second
y the inter-
which are
parison of
it rankings
>thod has a
purposes.
d truth in-
ods. They
isclassified
inal scales.
16
, M. 2005:
auflósende
ymposium,
ancing the
harpening.
New York).
g. Elsevier
age fusion
1s. Interna-
- 854, Mar.
nd spectral
mentation.
6(1), 49-61
ality Index.
2002
/software/-
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
A TRANSFORMATION METHOD FOR TEXTURE
FEATURE DESCRIPTION UNDER DIFFERENT IMAGINE CONDITIONS
Z. Guan, J. Yu, T. Feng , A. Li
Research Center of RS & Spatial Infor. Technology /Department of Surveying and Geo-informatics, College of Civil
Engineering, Tongji University
(zequnguan, 2011_jieyu,fengtiantian)@tongji.edu.cn
asia.aixia@gmail.com
Commission VII, Working Group VII/6
KEY WORDS: Transformation method, Texture feature, Gabor wavelet, Gaussian mixture models
ABSTRACT:
For high spatial resolution Remote Sensing images, it is very important to investigate the transformational methods between
background and target characteristics. Only in this way rich details in images under different imaging conditions can be well
extracted. Amongst the characteristics of imagery targets, texture is a visual feature that reflects the homogeneity of images and the
inner attributes of different objects. What’s more, it includes important information which describes the structural arrangement of
objects and the connection with the surrounding environment. This paper regards texture as the major feature and investigates the
transformational methods of texture feature description under different imaging conditions.
This paper mainly consists of three parts:(1) Construct a wavelet filter based on Gabor wavelet, which describes texture features
obtained under different imaging conditions;(2) Process and analyze the different object’s texture features jointly by the relationship
which is built by the wavelet description;(3) Build the transformation between the wavelet descriptions of the different object’s
texture features based on the characteristics of the band and direction.
1. INTRODUCTION
As a natural attribute of subjects, texture is a visual feature that
reflects the homogeneity of images. People have researched
image texture for more than 50 years and formed many methods
of texture features description under different imaging
conditions. There are several approaches to multispectral
texture description, both supervised and unsupervised. Haralick
(1973) presented a new method called gray level co-occurrence
matrix(GLCM) which is now widely used. He applied GLCM
to Landsat-1 multispectral image of the California coastal area
to solve the land use problem. Weszk et al. (1976) researched
texture for terrain analysis by using first-order statistics of gray
level differences and second-order gray level statistics. Lopez-
Espinoza et al. (2008) presented a method for image
classification which was taken by SPOT-5 and TM, based on
tree-structured Markov random field (TS-MRF) and a texture
energy function (TEF). Chellappa et al. (1985) applied
Gaussian Markov random field (GMRF) models to image
classification. Pentland (1984) put forward that fractals can be
used in the area of texture features description. Shu (1998)
described the SPOT image of Wuhan based on the method of
fractal assessment in image texture analysis. Chitre and Dhawan
(1999) used multi band wavelet for natural texture classification.
People began to research SAR image texture after the first radar
satellite running in earth orbit launched successfully by
America. Soh and Tsatsoulis (1999) used GLCM to analyze sea
ice texture with 100-m ERS-1 synthetic aperture radar (SAR)
imagery. Duskunovic et al. (2000) detected urban areas with the
Markov Random Field (MRF) texture classification in SAR
imagery. Hu et al. (2001) extracted texture information from
Radarsat imagery of Xuzhou with Daubechies3 orthogonal
wavelet successfully. Ni et al. (2004) used orthogonal wavelet
and second generation wavelet for SAR image classification
and compared the results from different methods with each
other. Ivanov and Paschenko (2006) studied SAR image
segmentation based on fractal dimension field. People also have
researched infrared imagery texture description recently. Song,
Wan and Chen (2006) applied GLCM to TM6 infrared imagery
for image enhancement by computing six textural features. Lin
et al. (2009) established the detection probability model based
on texture feature of thermal infrared image with Gabor wavelet.
All of the above methods of texture features description under
different imaging conditions can be divided into five categories
according to the principle proposed (Tuceryan, Jain, 1993).
They are statistical methods, geometrical methods, structural
methods, model based methods and signal processing methods.
Both GLCM and Gabor wavelet are the most popular methods
among all of them. And this paper will research the
investigating transformational methods of texture features
description under different imaging conditions based on Gabor
wavelet.
The wavelet theory has been utilized in image texture analysis
since it was introduced into the area of image process (Mallat,
1989). And among all branches developed from wavelet, Gabor
wavelet has been proved to be the optical filter of both spatial
domain and frequency domain under 2D uncertainty. Dunn,
Higgins and Wakeley (1994) devised a more rigorous method
for designing 2D Gabor filters and utilized it to segment images.
Wu et al. (2001) designed a optimal Gabor filter for Bi-textured
image segmentation with the Fourier power spectrum density.
Chen and Wang (2007) integrated Gabor wavelet and
independent component analysis (ICP) for image classification.
Clausi and Jernigan (2000) used Gabor filters to classify a SAR
aerial image and obtain textures from the Brodatz album based
on human visual system (HVS). Arivazhagan et al. (2006)
proposed a method of image classification using Gabor filter
based on rotation invariant features. Also Gabor wavelet has
been utilized in other areas. Lin et al. (2007) evaluated the
result of camouflage with Gabor wavelet. Song, Liu and Xie
