基于几何形状的彩色纹理分析方法研究

项目名称： 基于几何形状的彩色纹理分析方法研究

项目编号： No.U1504608

项目类型： 联合基金项目

立项/批准年度： 2016

项目学科： 计算机科学学科

项目作者： 李玉华

作者单位： 郑州轻工业学院

项目金额： 27万元

中文摘要： 纹理是自然图像结构的基本组成元素，是图像中普遍存在而又难以描述的特征。纹理分析一直是图像处理、计算机视觉等领域重点研究的基础内容之一。然而，由于纹理的复杂性以及概念的不明确性，使得纹理分析成为一个具有挑战性的课题，已有的方法存在着不同的局限性，在真实场景中的应用的结果都不太理想。本项目通过提取纹理的几何形状特征，研究更符合人类视觉感知的具有良好的几何不变性及对比度不变性的彩色纹理分析方法。主要包括：（1）彩色纹理图像的多尺度的、完备的图像表达方法，且基于几何形状并具有良好的对比度不变性；（2）平移、旋转、缩放不变的纹理几何特征的提取方法，使得基于该特征的纹理分析方法具备良好的几何不变性以及抗噪能力；（3）基于局部纹理几何特征的纹理图像的表达方法。这种纹理分析方法将在真实的织物图像检索平台上进行验证。相关研究成果可为图像处理、图像分析、计算机视觉等领域提供理论及技术支持。

中文关键词： 纹理分析；多尺度表达；几何不变；对比度不变性；视点不变性

英文摘要： As a fundamental ingredient of structure of natural image, texture widely exists in image and is hard to describe. Texture analysis is always one of research focuses in image processing and computer vision. However, it remains a challenging problem due to the high complexity and conceptual ambiguity of texture. Existing methods have different limitations so that the practical application of them is not ideal. In this project, we explore a color texture analysis method, which is invariant to geometric and contrast changes, and is more consistent with human visual perception by extracting the geometric characteristics of texture images. The content of this research includes several following aspects: (1) Geometric shapes based multi-scale and complete representation with invariance to contrast changes for color images; (2) the approach to extract texture features with invariances of translation, rotation and scale. This guarantees geometric invariance and anti-noise ability for the texture analysis methods using the approach; (3) local geometric features based texture image representation. The proposed method will be verified on a real world platform for fabric image retrieval. The outcomes of this project can provide theoretical and technological support for the domain of image processing, image analysis, and computer vision, and finally promote the development of related fields.

英文关键词： Texture Analysis;Muti-scale Representation;Geometric Invariance;Contrast Invariance;Viewpoint Invariance