05 Median Robust Extended Local Binary Pattern for Texture Classification

1.题目和关键词
Title: Median Robust Extended Local Binary Pattern for Texture Classification
基于中值鲁棒扩展局部二进模式的纹理分类
Keywords:
Texture descriptors纹理描述子;
rotation invariance旋转不变性;
local binary pattern (LBP)局部二进模式;
feature extraction特征提取;
texture analysis纹理分析.

2.摘要
Local binary patterns (LBP) are considered among the most computationally efficient high performance texture features. However, the LBP method is very sensitive to image noise and is unable to capture macrostructure information. To best address these disadvantages, in this paper, we introduce a novel descriptor for texture classification, the median robust extended LBP (MRELBP). Different from the traditional LBP and many LBP variants, MRELBP compares regional image medians rather than raw image intensities. A multiscale LBP type descriptor is computed by efficiently comparing image medians over a novel sampling scheme, which can capture both microstructure and macrostructure texture information. A comprehensive evaluation on benchmark data sets reveals MRELBP’s high performance robust to gray scale variations, rotation changes and noise—but at a low computational cost. MRELBP produces the best classification scores of 99.82%, 99.38%, and 99.77% on three popular Outex test suites. More importantly, MRELBP is shown to be highly robust to image noise, including Gaussian noise, Gaussian blur, salt and-pepper noise, and random pixel corruption.

局部二进模式(LBP)被认为是计算效率最高的高性能纹理特征之一。然而,LBP方法对图像噪声非常敏感,无法捕获宏观结构信息。为了更好地解决这些缺点,本文引入了一种新的纹理分类描述子,即中值鲁棒扩展局部二进模式(MRELBP)。与传统的LBP和大多数LBP变体不同,MRELBP比较的是区域图像中位数而不是原始图像强度。通过在一种新的采样方案有效地比较图像的中值,可以计算出多尺度LBP类型描述子,该方案可以捕获微观结构和宏观结构的纹理信息。通过对基准数据集的全面评估,发现MRELBP算法对灰度变化、旋转变化和噪声具有高鲁棒性,且计算成本较低。在三个主流的Outex测试套件上,MRELBP产生的最佳分类得分分别为99.82%,99.38%和99.77%。更重要的是,MRELBP被证明对图像噪声具有高度鲁棒性,其中包括高斯噪声,高斯模糊,椒盐噪声和随机像素损坏。

3.创新点、学术价值
基于ELBP方法存在对图像模糊和噪声敏感、未能捕捉纹理宏观结构及高特征维数的缺点,该论文提出了一种概念简单、质量高且计算效率高的方法,即基于中值滤波器(median filter)和多分辨率支集(multiresolution support)相结合的中值鲁棒扩展局部二进模式(MRELBP),创新点如下:
(1)该论文在DAISY,BRISK和FREAK的启发下,提出了一种新的采样方案,可以同时封装微观结构和宏观结构信息;
(2)该论文将局部中值与新的采样方案相结合发现能够形成显著的纹理特征。
(3)该论文从几个不同的角度对基准纹理数据集进行了全面评估,该方法包括采样参数,编码策略,光照不变性,旋转不变性,速度,分辨力和噪声鲁棒性。
(4)与10个基准纹理数据集上的11个最新的最新LBP变体相比,该方法具有灰度不变性(gray scale invariance)、旋转不变性(rotation invariance)、无需预训练和参数调整,并具有出色的分辨力(discriminativeness)和噪声鲁棒性。

4.对结论的理解和对学习工作的启发
We have presented a novel MRELBP descriptor to enhance the performance of current LBP variants. It outperforms recent state of the art LBP type descriptors in noise free situations and demonstrates striking robustness to image noise including Gaussian white noise, Gaussian blur, Salt-and-Pepper and pixel corruption. The proposed MRELBP has attractive properties of strong discriminativeness, gray scale and rotation invariance, no need for a pretraining, no tuning of parameters, and computational efficiency. As future work, we wish to investigate high–level applications such as image patching and object recognition.
提出了一种新的MRELBP描述子来提高现有LBP变体的性能。它在无噪声的情况下优于目前最先进的LBP类型描述子,并且对图像噪声(包括高斯白噪声、高斯模糊、椒盐噪声和像素损坏)具有显著的鲁棒性。该算法(MRELBP)具有显著的分辨力、灰度和旋转不变性、无需预训练、无需参数调整、计算效率高等优点。在未来的工作中,我们希望研究高水平的应用,如图像修补和目标识别。

猜你喜欢

转载自blog.csdn.net/weixin_37996254/article/details/108900355