The texture is an essential characteristic of the image. So, recognition of texture is increasingly becoming a major topic in many image processing applications such as image retrieving, image classification, similarity, object recognition, and detection. The recognition of texture tries to allocate an unidentified image to one of the identified class of textures. This paper proposes a novel feature extraction technique for classification and recognition of color texture. The significant advantage of the introduced method is that it combines the extraction of local and global features of the color texture by using Local Binary Pattern (LBP) and multi-channel orthogonal radial substituted Chebyshev moments, respectively. Relevant features (local or global) provides discriminatory information that used to differentiate one object from another. Global features represent the image as a whole, while local features represent a specific part of the image. We performed experiments using challenging datasets: (Outex, ALOT) to test the efficacy of our image classification descriptors. The result of this approach has said that our descriptor is valid, competitive, discriminatory, and exceeds the current state-of-art methods.

纹理是图像的基本特征。因此，在许多图像处理应用程序中，例如图像检索，图像分类，相似度，对象识别和检测，纹理识别正日益成为一个主要主题。纹理的识别尝试将未识别的图像分配给已识别的纹理类别之一。本文提出了一种新的特征提取技术，用于颜色纹理的分类和识别。引入的方法的显着优点是，它分别通过使用局部二值模式（LBP）和多通道正交径向替换的切比雪夫矩来组合颜色纹理的局部和全局特征的提取。相关功能（本地或全局）提供了用于将一个对象与另一个对象区分开的歧视性信息。全局特征代表整个图像，而局部特征代表图像的特定部分。我们使用具有挑战性的数据集进行了实验：（Outex，ALOT）以测试我们的图像分类描述符的功效。这种方法的结果表明，我们的描述符是有效的，具有竞争力的，歧视性的，并且超出了当前的最新方法。