遥感影像纹理分析方法综述与展望

遥感影像的纹理分析已经成为一种重要的提高遥感影像分类精度的手段。着重介绍了用于遥感影像纹理分析的方法,对这些方法进行了分类和综合;这些方法的类别是:统计方法、结构方法、模型方法以及基于数学变换的方法。接着分别对各类别中的多种纹理分析方法进行了剖析,列举各自的纹理特征,并指出了这些方法的优缺点和适应性。然后对应用这些方法的影像分类效果做了对比分析。最后分析了遥感影像纹理分析近年来的发展方向并对未来发展进行了展望。

Texture Analysis Methods Used In Remote Sensing Images

Image texture analysis has received a considerable amount of attention over the last few years as it has played an important role in the classification of the remote sensing images. Here introduced the texture analysis methods used for remote sensing images and made them categorised and sorted. The categories include: statistical method, structural method, model based method and mathematical transform method. Then it presented the texture analysis methods of every category, enumerated the texture feature respectively and introduced the advantage, disadvantage and adaptability of these methods. These methods include: grey level co-occurrence matrices method, the grey level run length method, autocorrelation method, Markov random field method, fractal model, power frequency method, wave-let transform method, first order histogram method, third order histogram method, grey level difference vectors, filters of space field, texture spectrum method, prospectrum, autoregressive model, binary stack method and logic operators. After that, it made comparison of classification performance via these methods. Finally, it analysed the trend of remote sensing texture analysis in the recent years and bring forward some prospects.