对反射光谱数据进行去噪是提高光谱信息准确度的前提。 传统时域平滑和频域去噪方法存在诸多缺点, 本文首次将广义形态滤波方法用于可见近红外光谱的去噪处理, 并提出基于小波包变换和数学形态学结合的光谱去噪方法。 使用USGS光谱库中的植被光谱进行实验, 采用信噪比(SNR)、 均方误差根(RMSE)、 波形相似度(NCC)和平滑度(SR)四个指标来评估去噪效果。 结果表明, 小波包最佳基阈值法和广义形态滤波法都能较好地保持波形和平滑度, 广义形态滤波法能较好地消除幅值较大的随机噪声, 但其对连续随机噪声中幅值较小的噪声成分不能有效消除； 而小波包最佳基阈值法不能有效消除幅值较大的噪声成分； 二者结合的方法组合了这两者的优点, 使得幅值较大、 较小的噪声成分都能较好地消除, 同时还提高了相似度和平滑度指标, 充分表明小波包最佳基阈值与广义形态滤波结合的方法是一种更好的可见光近红外光谱去噪方法。Wavelet Package Transform and Mathematical Morphology

The present study introduced the generalized morphological filter intothe denoising of visible and near infrared spectra for the first time, andprovided a new method for denoising the reflectance spectra by combiningmathematical morphology methods with the wavelet packet transformation. Theauthors used vegetable spectra from USGS spectral library as the referencespectra, and obtained the noised spectra by adding noises with different signal-to-noise ratios to the referenced spectra. The results were evaluated by signal-to-noise ratio (SNR), root mean squared error (RMSE), normalized correlationcoefficient (NCC) and smoothness ratio (SR) of the denoised spectra. The authors’ results showed that both the thresholding on wavelet packet decomposition bestbases method and the generalized morphological filter method could maintain thespectral shape and the spectral smoothness after denoising. The generalizedmorphological filter method can remove larger amplitude random noise whereas thecontinuous small amplitude random noise could not be removed well. Hence, thedenoised spectra were not smooth. Nevertheless, the denoised spectra using thethresholding on the best base groups of wavelet packet decomposition method weresmooth, but the larger amplitude noise could not be removed completely. Theauthors’ method by combining the two methods has the merits of the two methodsbut removing their defects. The results showed that both large and smallamplitude noise could be removed completely, meanwhile the normalized correlationcoefficient (NCC) and smoothness ratio (SR) were improved, which indicated thatthe authors’ method is superior to other methods in denoising visible and nearinfrared spectra.morphological filter; Thresholding on wavelet packet decomposition best bases与中国科学院创新项目(Kzcx2-yw-107)联合资助