新的纹理分类算法

马晓川; 侯朝焕; 唐姗

doi:10.11834/jig.19990591

学术论文与技术报告 | 浏览量 : 0 下载量: 173 CSCD: 0

PDF
导出
分享
收藏
专辑

新的纹理分类算法
Texture Classification Based on the Feature Symbol Random Field
1999年4卷第5期页码：387
纸质出版：1999
DOI： 10.11834/jig.19990591
稿件说明：

移动端阅览

马晓川, 侯朝焕, 唐姗. 新的纹理分类算法[J]. 中国图象图形学报, 1999,4(5):387. DOI： 10.11834/jig.19990591.

Texture Classification Based on the Feature Symbol Random Field[J]. Journal of Image and Graphics, 1999, 4(5): 387. DOI： 10.11834/jig.19990591.

摘要

提出了一种基于特征符号随机场描述的纹理分类方法。特征符号随机场是纹理图象的一种二维描述，综采用离散的符号来描述纹理图象的结构性特征。

Abstract

This paper suggests a texture classification algorithm based on the feature symbol random field (FSRF). FSRF is a 2 D representation of the image's bank information. In FSRF

symbol value of a pixel gives out its structure function in texture region and is more valuable than the gray value for texture classification. Meanwhile

several co occurrence matrix features are driven from the FSRF which are more powerful than energy features. This paper also suggests a Hierarchical scheme which leads to 96% correct ratio in the included experiments

while the gray value based method 67.5% and bank energies based method 75%.