ISSN 1054-6618, Pattern Recognition and Image Analysis, 2017, Vol. 27, No. 3, pp. 473–479. © Pleiades Publishing, Ltd., 2017.
Texture Classification Using Partial Differential Equation Approach
and Wavelet Transform
1
P. S. Hiremath
a
and Rohini A. Bhusnurmath
b,
*
a
Department of Computer Science (MCA), KLE Technological University, BVBCET Campus, Hubli-580031, Karnataka, India
b
Department of P.G. Studies and Research in Computer Science, Gulbarga University, Kalaburagi-585106, Karnataka, India
*e-mail: rohiniabmath@gmail.com
Abstract—Textures and patterns are the distinguishing characteristics of objects. Texture classification plays
fundamental role in computer vision and image processing applications. In this paper, texture classification
using PDE (partial differential equation) approach and wavelet transform is presented. The proposed method
uses wavelet transform to obtain the directional information of the image. A PDE for anisotropic diffusion is
employed to obtain texture component of the image. The feature set is obtained by computing different sta-
tistical features from the texture component. The linear discriminant analysis (LDA) enhances separability of
texture feature classes. The features obtained from LDA are class representatives. The proposed approach is
experimented on three gray scale texture datasets: VisTex, Kylberg, and Oulu. The classification accuracy of
the proposed method is evaluated using k-NN classifier. The experimental results show the effectiveness of
the proposed method as compared to the other methods in the literature.
Keywords: texture classification, partial differential equations, k-NN, computational cost, wavelet transform
DOI: 10.1134/S1054661817030154
1. INTRODUCTION
Texture is omnipresent in natural images. It con-
tains important characteristics which can be used in a
variety of image analysis application, which includes
image retrieval, image segmentation, and shape from
texture. Texture classification is a primary problem in
computer vision and image processing. It plays an
important role in a variety of applications such as
object recognition, medical image analysis, content-
based image retrieval, remote sensing. Feature
extraction is an important part in texture classifica-
tion, with extensive surveys [1–3]. For better classifi-
cation, the most discriminant features of various tex-
ture classes must be obtained from image. Recent
approaches for texture analysis are dependent on mul-
tiresolution analysis [9]. This property simulates
human visual system. The wavelet transform is popu-
lar multiresolution approach. The wavelet methods
have computational advantages compared to other
methods for texture classification [5, 18]. The method
for texture feature extraction using wavelet co-occur-
rence histograms is effectively applied to content
based image retrieval system [26] and script identifica-
tion [27]. Several works [6] shows that single texture
descriptor cannot bring enough information for obtain-
ing good results in datasets. Randen and Husøy [3] con-
cluded in their study of dozens of different filtering
1
The article is published in the original.
methods: “No single approach did perform best or very
close to the best for all images; thus, no single approach
may be selected as the clear winner of this study.”
In [7], texture analysis using the nonsubsampled
contourlet transform (NSCT) and local directional
binary patterns (LDBP) and k-NN classifier for classi-
fication is proposed. Shift and rotation invariant texture
classification using support vector machine is intro-
duced in [8]. Local directional binary patterns in texture
image are taken as texture descriptors. The dominant
LDBPs are explored for different neighborhoods of dif-
ferent radial distances of small kernel. These are the
basic properties of local image texture [9]. Images are
represented using multiple scales by decomposing an
image into set of derived images [10, 11]. Perona and
Malik [12] introduced the concept of inter region
smoothing and keeping the edges sharp, called aniso-
tropic diffusion (AD). A texture classification method
using AD and LDBP is experimented on Brodatz data-
set in [13]. Texture image classification for RGB colour
space using partial differential equation (PDE) method
is experimented on Oulu colour dataset in [14]. Texture
analysis based on PDE and local directional binary pat-
terns is tested on different datasets in [15].
The present study is the extension of authors previ-
ous work [16], wherein the texture descriptors are
obtained using partial differential equation and wave-
let transform, that are tested on Brodatz dataset for
classification. The objective of study is to investigate
the effectiveness of the proposed method on VisTex,
Oulu and Kylberg gray scale datasets to establish the
robustness of the method. These datasets differ in num-
REPRESENTATION, PROCESSING, ANALYSIS,
AND UNDERSTANDING OF IMAGES
Received May 10, 2016