基于特征尺度及多尺度分解的纹理分割

在计算机视觉领域，纹理识别及分割是图像低层次处理中一个重要的问题。文章依据视觉信息处理的生理学理论，提出了一种新的纹理分析算法。该算法通过搜寻纹理图像的特征尺度并配合连续整数尺度滤波上的特征尺度统计对纹理进行分层分析。并基于特征尺度上的分层分析提出一种纹理分割算法。实验结果证明了该算法的有效性。

Characteristic Scale and Multiscale Analysis-based Texture Segmentation

The identification and segmentation of texture is an important low-l evel operation in computer vision that has many applications. This paper propos e s a new texture analysis algorithm which is based on the theory of visual info rm ation processing in physiology. The algorithm utilizes searching texture char act eristic scale and extracting the statistic characteristics on the texture ch arac teristic scale after continuous integer scale filtering to analyze the prop erty of texture. Furthermore, a texture segmentation algorithm based on the cha racter istic scale of texture is proposed in this paper. Promising experimental results show the efficiency of the algorithm.