Abstract

The issues of joint estimation of the scaling, singular, multifractal, and anisotropic properties of image texture are considered. A technique for forming a fundamentally new class of topological textural-multifractal descriptors for joint estimation of various fractal properties of a texture is described. The efficiency of segmentation of real radar and optical images by the Fuzzy c-means clustering algorithms in the space of the proposed class of descriptors is estimated. Based on the results and their comparison with the accuracy of segmentation by the existing descriptor, the conclusion about the higher informativity of the new class of topological fractal descriptors is made.

中文翻译：

---

一类新的拓扑纹理多重分形描述符及其在处理低对比度雷达和光学图像中的应用

摘要

考虑了图像纹理的尺度、奇异、多重分形和各向异性属性的联合估计问题。描述了一种用于形成一种全新的拓扑纹理多重分形描述符的技术，用于联合估计纹理的各种分形属性。估计了在所提出的描述符类别的空间中通过模糊 c 均值聚类算法分割真实雷达和光学图像的效率。基于结果及其与现有描述符分割精度的比较，得出了新的拓扑分形描述符具有更高信息性的结论。