一組基于廣義局部沃爾什變換的紋理特征
A Group of Texture Features Based on Generalized Local Walsh Transform
-
摘要: 該文提出一組基于廣義局部沃爾什變換(GLWT)的紋理特征。首先給出局部沃爾什變換(LWT)的定義，并在空域中對其加以推廣，用以提取圖像的局部紋理信息；然后在一個宏窗口中估計12個GLWT系數(shù)的二階矩作為圖像的紋理特征。對這組紋理特征的鑒別性能進行了分析，并與Haralick(1973)，Wang Li(1990)，以及Yu Hui提出的紋理特征進行了比較。實驗結(jié)果表明，該文提出的紋理特征具有更好的鑒別性能和分類能力。
-
關(guān)鍵詞:
- 圖像處理;模式識別;廣義局部沃爾什變換;紋理特征
Abstract: A new group of texture features based on Generalized Local Walsh Transform (GLWT) are presented in this paper. Firstly, Local Walsh Transform (LWT) is defined and generalized in spatial domain to extract local texture information. Then, the twelve 2nd order moments of GLWT coefficients are estimated in a macro-window and selected as texture features. The performance of these features in discriminating textures has been evaluated. The textural features from Haralick(1973), Wang Li(1990), Yu Hui has been used for comparison. The experimental results reveal that the texture features presented have the best discriminating performance. -
Haralick R M, Shanmugam K, DinStein I. Texture features forimage classification. IEEE Tran,s. on Systems Man Cybernet,1973,SMC-3 (6): 610-621.[2]Wang Li, He Dongchen. Texture classification using texturespectrum[J].Pattern Recognition.1990, 23(5):905-910[3]He Dongchen, Wang Li. Texture features based on texturespectrum[J].Pattern Recognition.1991, 24(5):391-399[4]Zhou F, Feng J F, Shi Q Y. Image segmentation based on localFourier coefficientsMultispectral Imagehistogram.ProcessingProc. SPTF-2nd Tnt. Conf. onand Pattern Recognition, Wuhan,China, November, 2001: 143-152.[5] Yu Hui, Li Mingjing, Zhang Hongjiang, Feng Jufu. Color texturemoments for content based image retrieval. www.cs.iupui.edu/}tuceryan/research/ComputerVision/moment-paper.pdf.[6]孫即祥.數(shù)字圖像處理[M].鄭州:河南教育出版社，1993:99-105.[7]Simona E, Crigeorescu, Petkov Nicolai, Kruizinga Peter.Comparison of texture features based on Cabor filters[J].IEEbTrans. on Image Processing.2002, 11(10):1160-[8]孫即祥等.現(xiàn)代模式識別「M].長沙:國防科學(xué)技術(shù)大學(xué)出版社，2002: 31-36. -
計量
- 文章訪問數(shù): 2573
- HTML全文瀏覽量: 104
- PDF下載量: 1151
- 被引次數(shù): 0
下載:
