Texture image retrieval using phase-based features in the complex wavelet domain

Texture Image Retrieval using Phase-Based Features

in the Complex Wavelet Domain

Hoang Nguyen-Duc

Research & Development Department

Broadcast Research and Application Center

Ho Chi Minh City, Vietnam

[email protected]

Tuan Do-Hong

Electric-Electronic Department,

Ho Chi Minh University of Technology

Ho Chi Minh City, Vietnam

[email protected]

Thuong Le-Tien

Electric-Electronic Department,

Ho Chi Minh University of Technology

Ho Chi Minh City, Vietnam

[email protected]

Cao Bui-Thu

Electronic Telecommunication Division

Ho Chi Minh City University of Industry

Ho Chi Minh City, Vietnam

[email protected]

Abstract—The phase holds crucial information about image

structures and features, but only the real part or the magnitude

of the transform coefficients is often used for image processing

applications. In this paper, a method for the feature extraction of

images called Phase-based LBP is presented. Proposed method is

based on the combination of phase of complex wavelet

coefficients and the Local Binary Pattern operator (LBP). We

also perform the comparative analysis about retrieval

effectiveness of phase information of some complex wavelet

transforms using for Phase-based LBP. Experimental results,

achieved with the standard rotated Brodatz dataset, show the

interest of this method comparing with another methods only

based on the real part or the magnitude of the wavelet

coefficients for texture image retrieval.

Keywords-complex wavelet; phase information; steerable

pyramid; Local Binary Pattern; LBP; image retrieval.

I. INTRODUCTION

Wavelet and filter banks have been studied in image

processing applications in a long time. In wavelet-based image

retrieval, many researchers focus on extracting information

from real coefficients of wavelet transforms of the next

generation as contourlet [4], curvelet [8],… or the magnitude of

coefficients from complex wavelet transforms [5]. References

[4, 8, 5, 12] also have been demonstrated the texture image

retrieval based on only the real part or the magnitude of the

transform coefficients that have retrieval effectiveness.

Several applications exploit the phase information across

scales of the complex wavelet transforms such as the

description of texture image in [10], the investigation of local

phase based on the dual-tree complex wavelet transform [13]

and the complex directional filter bank (CDFB) [3], modeling

natural images by the probability density function of relative

phase [2]… Therefore, the phase of the complex wavelet

coefficients can be beneficial to the development in the image

retrieval application.

Reference [14] proposed to use the Local Binary Pattern

(LBP) operator for rotation invariant texture classification.

Some important major characteristics of this operator include

its computational efficiency as well as its high discriminative

properties at local regions. The LBP operator has achieved

impressive classification results on representative texture

databases [15] and have been used to many other applications,

such as face recognition [16], dynamic texture recognition

[6],…

Combination of the wavelet transform or the Fourier

transform with the LBP operator have been used in some works

such as [17, 9, 18]. In [17], Local Binary Pattern Histogram

Fourier features (LBP-HF) are proposed from the discrete

Fourier transform of LBP histograms. This is a rotation

invariant image descriptor based on uniform LBP [14]. With

[9], texture images are characterized by exploiting the

multiresolution properties of the Steerable Pyramid

Decomposition, and features are extracted from magnitudes of

steerable multiresolution subbands by applying the LBP

operator. Reference [18] proposes an image representation

method for face recognition called local Gabor phase

difference pattern (LGPDP). The LGPDP captures the Gabor

phase difference relationships to represent an image. Gabor

phase differences between the center pixel and neighborhoods

are all calculated for each in image (the same with the

calculated method of the LBP operator but coding rule is

difference).

This paper attempts to propose a generalized method to

extract features of texture images based on phase-based

information in the complex wavelet domain. First, a texture

image is decomposed by a complex wavelet transform; then

phase information of complex wavelet coefficients are

extracted at each scale and orientation; finally, the LBP

operator (using the proposed coding rule) is applied on each

subband to build feature vectors of texture images. Our method

The 2010 International Conference on Advanced Technologies for Communications 978-1-4244-8873-5/10/$26.00 ©2010 IEEE