机构地区: 吉林大学
出 处: 《长春理工大学学报(自然科学版)》 2011年第3期132-134,155,共4页
摘 要: 当前常用的纹理特征描述方法多是以统计学为基础的,主要缺陷在于其纹理特征的提取尺度比较单一。为了弥补这个不足,本文提出了一种模型化了的统计的多分辨无监督纹理分割算法,并基于灰度层共生矩阵使之得以实现。实验结果表明在错误识别率上基于灰度层共生矩阵的一般的分割算法比本文所的述算法要高。 This paper presents a model of statistical multi-resolution unsupervised texture segmentation algorithm and re-alization based on the co-occurrence matrix to improve the deficiency that some usually used statistical texture-based Feature Description methods.The main deficiency is that some of the usually used methods only extracted the texture feature in a single scale.The experiment results show that the advantage of this algorithm is it's lower in error recog-nition rate than the usually used methods.
领 域: [自动化与计算机技术] [自动化与计算机技术]