王 旭,李小舟,雷慶鋒,李艷杰
(遼寧林業(yè)職業(yè)技術(shù)學(xué)院,遼寧沈陽 110101)
改進小波閾值法在木材圖像缺陷處理中的應(yīng)用
王 旭,李小舟,雷慶鋒,李艷杰
(遼寧林業(yè)職業(yè)技術(shù)學(xué)院,遼寧沈陽 110101)
閾值的選取和小波函數(shù)的構(gòu)造是小波閾值去噪的關(guān)鍵,但傳統(tǒng)的閾值去噪方法并沒有對此進行充分的研究。針對傳統(tǒng)小波閾值去噪分析的不足,分別從2個方面進行改進。首先,根據(jù)小波系數(shù)在各尺度上的相關(guān)性,提出了基于Lip指數(shù)閾值尋優(yōu)的新方法;其次,用構(gòu)造出新的小波閾值函數(shù)處理小波系數(shù),克服了軟閾值的高階不可導(dǎo)和硬閾值函數(shù)的不連續(xù),且計算方便。最后,對帶有節(jié)子裂縫的木材圖像進行去噪處理。實例驗證了改進小波閾值法的優(yōu)越性和有效性。
木材缺陷;小波去噪;閾值函數(shù);圖像處理;均方誤差
Abstract:Threshold selection and wavelet function construction were two key points of wavelet threshold denoising,which were not researched adequately in traditional threshold denoising methods.Deficiencies of traditional wavelet threshold methods in denoising were improved in two aspects.Firstly,according to the correlations of wavelet coefficients in different scales,new method for searching optimal threshold were proposed based on Lip exponent;secondly,new wavelet function was constructed to deal with wavelet coefficients,which not only convenient in calculating but also could overcome the non-differentiability in higher order of soft threshold and discontinuity of hard thresholding function;lastly,denoising treatment was done to the images of wood with knots and cracks.The superiority and effectiveness of improved wavelet thresholding method were tested.
KeyWords:wood defects;wavelet denoising;threshold function;picture processing;mean squared error
節(jié)子是樹木在生長過程中最普遍存在的一種現(xiàn)象。它不僅破壞木材紋理構(gòu)造的完整性和均勻性[1-2],使紋理質(zhì)量下降,大大降低木材的檔次,而且會給木材加工工藝帶來一些困難,降低木材的利用效率。由于木材的裂縫節(jié)子的像素值和木材背景顏色相近,邊緣也不清晰,會造成木材紋理混淆、分析困難等問題,因此必須要對帶有節(jié)子裂縫的木材圖像進行去噪處理,將節(jié)子裂縫部分最大平滑的同時保護紋理的細節(jié)部分[3],改善木材強度性質(zhì),提高木材加工工藝的利用率。
小波閾值去噪法因其具有多分辨率分析的特點,已被廣泛應(yīng)用于木材圖像去噪處理中[4-5],但是傳統(tǒng)的小波閾值去噪能力有限,所以很多學(xué)者提出了相應(yīng)的改進方法。……