基于概念的自然纹理分类

Classification of Natural Textures Based on Concepts

纹理是图像的重要视觉特征,纹理分类是图像分析、计算机视觉等领域一个重要的研究课题。文章不同于以往的纹理分类方法,提出了一种基于概念的纹理分类方法。该方法以中文自然语言中常用的纹理描述词作为纹理概念,给出了10个基本概念的纹理分类,然后利用Gabor滤波参数和SVM对自然纹理图像进行分类,实现了图像的纹理视觉特征到纹理概念的转换,部分解决了纹理概念与纹理参数之间的“语义鸿沟”问题。

Texture is an important feature of an image,therefore texture classification is an interesting topic of research on image analysis and computer vision.Distinguishing from the existing approaches of texture classification,a novel approach is proposed to classify natural textures based on the concepts which are denoted by some familiar Chinese words describing various natural textures.Ten concepts are presented to automatically classify natural textures with Gabor filter parameters and SVM to transform the texture parameters to texture concepts.The work is useful to negotiating the ＂semantic gaps＂ between texture concepts and feature parameters on image understanding and image retrieval based on natural language.