CLBP算法：一种局部二值模式的完整建模及其在纹理分类中的应用

1星需积分: 9 72 浏览量更新于2024-09-14 1 收藏 116KB PDF 举报

"本文提出了一种局部二值模式（LBP）操作的完整建模方法，并发展了关联的完整LBP（CLBP）方案，用于纹理分类。通过局部差异符号幅度变换（LDSMT），一个局部区域由其中心像素表示，并转化为CLBP-Center（CLBP_C）的二进制码。LDSMT将图像的局部差异分解为两个互补组件：符号和幅度，进而提出了CLBP-Sign（CLBP_S）和CLBP-Magnitude（CLBP_M）两个操作器来编码它们。传统的LBP相当于CLBP中的CLBP_S部分，研究表明CLBP_S比CLBP_M保留了更多的局部结构信息，解释了为什么简单的LBP操作器可以较好地提取纹理特征。通过联合或混合CLBP_S、CLBP_M和CLBP_C的特征分布，可以显著提高旋转不变纹理分类的效果。关键词包括：局部二值模式，旋转不变性，纹理分类，CLBP，LDSMT。" 文章详细说明了CLBP在纹理分类中的应用和改进。LBP是一种常用的纹理分析和特征提取方法，它通过比较像素邻域内的灰度差异并将这些差异转化为二进制码来表达纹理特征。然而，原始的LBP方法在处理旋转变化的纹理时存在不足。为了解决这一问题，文章提出了CLBP，一种更全面的LBP模型。 CLBP的核心是通过LDSMT来表征图像的局部差异。LDSMT将图像的局部差异分解为两个方面：符号和幅度。CLBP_S编码的是符号信息，而CLBP_M编码的是幅度信息。作者指出，CLBP_S保留了更多关于局部结构的信息，这是原始LBP能够成功提取纹理特征的关键原因。同时，CLBP_M虽然包含幅度信息，但在保持旋转不变性方面可能略逊一筹。为了进一步提升分类性能，研究者将CLBP_S、CLBP_M和CLBP_C的特征结合，创建了联合或混合分布。这种策略允许模型更好地捕捉纹理的多样性和复杂性，从而在旋转不变纹理分类任务中实现显著的性能提升。该论文提出的CLBP方法不仅扩展了LBP的理论框架，还提供了一个更强大的工具来处理纹理分类问题，尤其是在应对旋转变化时。这对于图像分析和计算机视觉领域有着重要的实际应用价值，特别是在人脸识别和其他依赖纹理特征识别的任务中。

Submitted to IEEE Transactions on Image Processing



Abstract—In this paper, a completed modeling of the LBP

operator is proposed and an associated completed LBP (CLBP)

scheme is developed for texture classification. A local region is

represented by its center pixel and a local difference

sign-magnitude transform (LDSMT). The center pixels represent

the image gray level and they are converted into a binary code,

namely CLBP-Center (CLBP_C), by global thresholding.

LDSMT decomposes the image local differences into two

complementary components: the signs and the magnitudes, and

two operators, namely CLBP-Sign (CLBP_S) and

CLBP-Magnitude (CLBP_M), are proposed to code them. The

traditional LBP is equivalent to the CLBP_S part of CLBP, and

we show that CLBP_S preserves more information of the local

structure than CLBP_M, which explains why the simple LBP

operator can extract the texture features reasonably well. By

combining CLBP_S, CLBP_M, and CLBP_C features into joint

or hybrid distributions, significant improvement can be made for

rotation invariant texture classification.

Index Terms—Local Binary Pattern, Rotation Invariance,

Texture Classification

I. INTRODUCTION

exture classification is an active research topic in computer

vision and pattern recognition. Early texture classification

methods focus on the statistical analysis of texture images. The

representative ones include the co-occurrence matrix method

[1] and the filtering based methods [2]. Kashyap and

Khotanzad [3] were among the first researchers to study

rotation-invariant texture classification by using a circular

autoregressive model. In the early stage, many models were

explored to study rotation invariance for texture classification,

including hidden Markov model [4] and Gaussian Markov

random filed [5]. Recently, Varma and Zisserman [6] proposed

to learn a rotation invariant texton dictionary from a training set,

and then classify the texture image based on its texton

distribution. Later, Varma and Zisserman [7-8] proposed

another texton based algorithm by using the image local patch

to represent features directly. Some works have been recently

proposed for scale and affine invariant texture classification by

using fractal analysis [9-10] and affine adaption [11-12].

In [13], Ojala et al proposed to use the Local Binary Pattern

(LBP) histogram for rotation invariant texture classification.

LBP is a simple yet efficient operator to describe local image

pattern, and it has achieved impressive classification results on

The work is partially supported by the GRF fund from the HKSAR

Government, the central fund from Hong Kong Polytechnic University, and the

National Science Foundation of China.

Z. Guo, L. Zhang and D. Zhang are with Department of Computing, the

Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong, China.

({cszguo, cslzhang, csdzhang}@comp.polyu.edu.hk).

* Corresponding author, phone: 852-27667271, fax: 852-27740842

representative texture databases [14]. LBP has also been

adapted to many other applications, such as face recognition

[15], dynamic texture recognition [16] and shape localization

[17].

Despite the great success of LBP in computer vision and

pattern recognition, its underlying working mechanism still

needs more investigation. Before proposing LBP, Ojala et al

[18] used the Absolute Gray Level Difference (AGLD)

between a pixel and its neighbours to generate textons, and

used the histogram of them to represent the image. Later, they

proposed LBP [13] to use the sign, instead of magnitude, of the

difference to represent the local pattern. Ojala et al [19] also

proposed a multidimensional distribution of Signed Gray Level

Difference (SGLD) and regarded LBP as a simplified operator

of SGLD by keeping sign patterns only. Ahonen and

Pietikäinen [20] analyzed LBP from a viewpoint of operator

implementation. Tan and Triggs [21] proposed Local Ternary

Pattern (LTP) to quantize the difference between a pixel and its

neighbours into three levels. Although some variants of LBP,

such as derivative-based LBP [17], dominant LBP [22] and

center-symmetric LBP [23], have been proposed recently, there

still remain some questions to be better answered for LBP. For

example, why the simple LBP code could convey so much

discriminant information of the local structure? What kind of

information is missed in LBP code, and how to effectively

represent the missing information in the LBP style so that better

texture classification can be achieved?

This paper attempts to address these questions by proposing

a new local feature extractor to generalize and complete LBP,

and we name the proposed method completed LBP (CLBP). In

CLBP, a local region is represented by its center pixel and a

local difference sign-magnitude transform (LDSMT). The

center pixel is simply coded by a binary code after global

thresholding, and the binary map is named as CLBP_Center

(CLBP_C). The LDSMT decomposes the image local structure

into two complementary components: the difference signs and

the difference magnitudes. Then two operators, CLBP-Sign

(CLBP_S) and CLBP-Magnitude (CLBP_M), are proposed to

code them. All the three code maps, CLBP_C, CLBP_S and

CLBP_M, are in binary format so that they can be readily

combined to form the final CLBP histogram. The CLBP could

achieve much better rotation invariant texture classification

results than conventional LBP based schemes.

Several observations can be made for CLBP. First, LBP is a

special case of CLBP by using only CLBP_S. Second, we will

show that the sign component preserves more image local

structural information than the magnitude component. This

explains why the simple LBP (i.e. CLBP_S) operator works

much better than CLBP_M for texture classification. Third, the

proposed CLBP_S, CLBP_M and CLBP_C code maps have

Zhenhua Guo, Lei Zhang, Member, IEEE, and David Zhang

, Fellow, IEEE

A Completed Modeling of Local Binary Pattern

Operator for Texture Classification

下载后可阅读完整内容，剩余6页未读，立即下载

canniao891030

粉丝: 0
资源: 3

CLBP算法：一种局部二值模式的完整建模及其在纹理分类中的应用

local binary pattern

CJLBP.rar_Completed lbp_noticewu3_silver72s

A completed version of XML bible

Teacher-completed child behavior checklist ratings as a function of classroom-based interventions: A pilot study

PhotoWall:Just sort out the photos and generate the photo wall with one click; Realize the alignment of a large number of photos of any scale without stretching and masking the photos; Completed by JS CSS only, independent of node or server.只需整理好照片，一键生成照

Because this call is not awaited, execution of the current method continues before the call is completed. Consider applying the 'await' operator to the result of the call.

Module load completed but symbols could not be loaded for dxgkrnl.sys

最新资源