纹理是在图像的局部区域内表现出不规则性，但是在整体上却表现出了某种规律

The texture shows irregularities in the local area of ​​the image, but shows a certain regular feature as a whole. It does not depend on the change of color and brightness, and reflects the internal characteristics shared by the surface of the object. The characteristics of scale, hierarchy, translation invariance and duality of certainty and randomness. <br>Texture feature extraction methods are: statistical method, structural method, model method and spectrum method. The statistical method mainly uses the local correlation between pixels to describe the texture. It is suitable for analyzing delicate and irregular objects such as mountains and forests. Typical methods are gray-level co-occurrence matrices; the structural method uses the arrangement regularity of primitives to analyze Texture, suitable for large and regular objects with thousands of texture primitives, such as model method, tree grammar, etc .; model method uses the parameters of the model as texture features, such as random field model method; spectrum method is to treat the texture image as Dimensional signals, using filtering methods for analysis, mainly including Fourier transform, Gabor transform and wavelet transform, etc.

纹理是在图像的局部区域内表现出不规则性，但是在整体上却表现出了某种规律性的特征，它不依赖于颜色和亮度的变化，反映了物体表面所共有的内在特性，具有尺度性、层次性、平移不变性以及确定与随机二重性的特点。<br>纹理特征的提取方法有：统计法、结构法、模型法和频谱法。统计法主要利用像素间的局部相关性来刻画纹理，适用于分析如山脉、森林等纹理细腻而且不规则的物体， 典型的方法如灰度共生矩阵；结构法利用基元的排列规则性来分析纹理，适用千纹理基元较大并且有规则的物体，如图模型法、树文法等；模型法采用模型的参数作为纹理特征，如随机场模型法；频谱法是将纹理图像看成是二维信号，使用滤波的方法进行分析， 主要包括傅立叶变换、 Gabor 变换和小波变换等

Texture is irregular in the local area of the image, but on the whole, it shows some regular characteristics. It does not depend on the change of color and brightness, reflects the internal characteristics of the object surface, and has the characteristics of scale, hierarchy, translation invariance, determination and random duality.<br>The methods of texture feature extraction include statistical method, structural method, model method and spectrum method. The statistical method mainly uses the local correlation between pixels to describe the texture, which is suitable for analyzing fine and irregular objects such as mountains and forests, Typical methods such as gray level co-occurrence matrix; structural method uses the arrangement regularity of primitive elements to analyze texture, which is suitable for objects with large and regular texture primitive elements, such as graph model method and tree grammar; model method uses the parameters of model as texture features, such as random field model method; spectrum method regards texture image as two-dimensional signal, and uses filtering method for analysis, It mainly includes Fourier transform, Gabor transform and wavelet transform<br>