基于LWT的纹理特征提取方法.pdfVIP

27 3 JOURNAL OF NATIONAL UNIVERSITY OF DEFENSE TECHNOLOGY Vol. 27 No. 3 2005 : 1001- 2486(2005) 03- 0086- 06 LWT 张志龙, 鲁新平, 沈振康, 李吉成 ( ATR , 410073) : ( LWT) , LWT , LWT : LWT , , , LWT ( 2 [ 1] [2, 3] [ 5] 46 ) Haralick Wang He , H i Y , LWT , : ; ; ; ; :TP391 :A A Texture Feature Extraction Method Based on Local Walsh Transform ZHANG Zhi long, LU Xinping, SHEN Zhenkang, LI Jicheng (ATR Key Lab, National Univ. of Defense Technology, Changsha 410073, China) Abstract:A new text re feat re extraction method sing Local Walsh Transform ( LWT) is presented. The definition of LWT is given. The statistical properties of LWT coefficients are analyzed. The text re discrimination performance of the moments of LWT coefficients are investigated. Detail examinations reveal that the LWT coefficients of the nat ral text re images s ally do not yield to Ga ss distrib tion, their evenorder moments have high text re discrimination performance, while their oddorder moments have low nd th th text re discrimination performance. Hence, the evenorder (2 , 4 , 6 order) moments of the LWT coefficients are selected as text re feat res. Comparedwith the other text re feat res defined by Haralick[ 1] , Wang and He[2, 3] , H i Y [ 5] , the text re feat res we present have the best text re discrimination performance. Key words:Walsh transform; text re feat res; text re analysis; image processing; pattern recognition [4] 33 , , , , Wang He , [ 2, 3] ,