基于相容粒度空间模型的图像纹理识别

Image texture recognition based on tolerance granular space

信息粒广泛存在于现实世界中,是对现实的抽象。信息粒化旨在建立基于外部世界的有效的并以用户为中心的概念,同时简化人们对物理世界和虚拟世界的认识。使用信息粒进行计算是粒度计算的基本思想,在问题描述和求解中都具有重要的意义。主要研究一种新的粒度计算模型—相容粒度空间模型在图像纹理识别中的应用。该模型基于这样的思想,即人类具有根据具体的任务特性把相关数据和知识泛化或者特化成不同程度、不同大小的粒的能力,并进一步根据这些粒和粒之间的关系进行问题求解的能力。通过在多个数据集上和10种其他图像纹理识别算法的比较实验表明,该方法在图像纹理识别上具有良好的应用效果。

Information granules exist widely in the real world, and they are the abstraction of the reality. The granulation is to construct the concepts depending on the context effectively and is user oriented. Besides, it is also for the simplification of our understanding on the physical and virtual world. Using information granules to calculate is the basic concept of granular computing, and it is of great importance for problem description and solution. This paper mainly explores the application of tolerance granular space on image texture recognition. The basic idea of tolerance granular space is based on the human ability that is, people can abstract or synthesize the knowledge and data related to special tasks to different degrees or sizes granules, and accomplish the tasks with the help of the granules and relations among them. With some experiments, the effectiveness and efficiency of the image texture recognition methods were testified, and the application also indicates that the tolerance granular space is an effective tool to solve some problems in image processing fields.