图像纹理分类方法研究进展和展望

Texture Classification:State-of-the-art Methods and Prospects

纹理分类是计算机视觉和模式识别领域的一个重要的基本问题,也是图像分割、物体识别、场景理解等其他视觉任务的基础.本文从纹理分类问题的基本定义出发,首先,对纹理分类研究中存在的困难与挑战进行阐述;接下来,对纹理分类方面的典型数据库进行全面梳理和总结;然后,对近期的纹理特征提取方法的发展和现状进行归类总结,并对主流纹理特征提取方法进行了详细的阐述和评述;最后,对纹理分类发展方向进行思考和讨论.

Texture is a fundamental characteristic of many types of images. Texture classification is one of the essential tasks in the field of computer vision and pattern recognition. It is also the basis of other complex vision tasks, such as image segmentation, object recognition and scene understanding. In this paper, we first address the importance of texture classification and summarize the difficulties and challenges in the development of texture feature extraction approaches.Then we discuss the existing texture databases which are generally acknowledged as public evaluation bases for texture classification methods. Next, we review recent achievements in the study of texture feature development and provid detail discussion on prominent texture feature descriptors. Finally, we point out the future directions of texture classification.