589.txt

来自「This complete matlab for neural network」· 文本 代码 · 共 32 行

发信人: GzLi (笑梨), 信区: DataMining
标  题: Re: 请问Gabor
发信站: 南京大学小百合站 (Sat Aug 10 00:35:58 2002), 站内信件

gabor feature:
http://ams.egeo.sai.jrc.it/eurostat/Lot16-SUPCOM95/node17.html

The statistical features mainly give information on intensity variations 
of pixels, whereas texture features based on SGLD matrices characterise spatial
 geometrical regularities of pixels within a certain neighbourhood. A third
 important consideration in the selection of the basic features is related
 to the compactness of signal representation. In many cases, as in land cover
/land use classification task, the images are generally highly constrained
, i.e., self-correlated. The high degree of self-correlation in images arises
 from the fact that the objects to be classified tend to have morphological
 consistency, including first order correlations (i.e., locally similar lumin
ance
 values), second order correlations (i.e., edge continuations), as well as
 higher order correlations (i.e, homogeneous textures). To capture this corre
lation
 the distribution of the features should also be constrained, so that few
 of the features are strongly present in the signal representation. This 
motivates the use of Gabor filters (or functions) [25], which can provide
 a compact representation for real signals [12]. 

--
              ***  端庄厚重 谦卑含容 事有归着 心存济物  ***
今天你挖了吗？ DataMining  http://DataMining.bbs.lilybbs.net 
 MathToolshttp://bbs.sjtu.edu.cn/cgi-bin/bbsdoc?board=MathTools 

※ 来源:．南京大学小百合站 bbs.nju.edu.cn．[FROM: 211.80.38.29]