小波域非局部均值无芒隐子草叶切片盲去噪
|
摘要
在获取无芒隐子草叶切片图像时不可避免受到噪声的污染,易导致后续提取和测量特征参数的不准确。对于自然图像,事先并不知道其所含噪声的类型和方差,因而首先利用小波变换和曲线拟合确定切片图像噪声类型和强度;在此基础上,分别应用小波阈值去噪、非局部均值去噪和提出的非局部均值滤波(NLM)与小波阈值去噪相结合的方法对无芒隐子草叶切片图像进行去噪。实验结果表明:获取的切片图像噪声类型为高斯加性噪声,标准差为σ∈[1. 5,3. 5],用高斯函数对随机选取的10幅切片图像的高频HH子带能量分布进行拟合,拟合优度为R2=0. 990 7;用3种方法对含不同噪声大小的切片图像进行去噪,当噪声标准差为σ∈[1. 5,8]时,应用Beyes Shrink法去噪后,图像的峰值信噪比提高了3 d B,而NLM和本文提出的算法不适用;当噪声标准差为σ∈[8,15]时,NLM算法和提出的算法去噪效果相当,去噪后图像峰值信噪比提高了7. 5d B,应用Beyes Shrink算法提高了6. 5 d B;而当σ∈[15,30]时,使用提出的算法表现出较大的优越性,去噪后图像峰值信噪比提高了10. 53d B,是NLM算法的1. 4倍、Beyes Shrink法的1. 3倍。本文的算法和实验结论可为无芒隐子草切片图像准确降噪提供理论基础。
The microscopic images of Cleistogenes Songorica's leaf are inevitably corrupted by noise in obtaining,which will be easy to lead to the inaccuracy of its subsequent extraction and measurement of the feature parameters. Wavelet transform and curve fitting are used to determine the type and intensity of the microscopic image noise,in this paper. On this basis,wavelet threshold method,non local algorithm and the proposed combined non local mean filter( NLM) and wavelet threshold method are applied to de noise the slice images. The results show that: The noise type was recognized as additive Gaussian noise. Energy distribution of high frequency HH sub-band of 10 microscopic images randomly selected were fitted with the Gaussian function,with the R2 value of 0. 990 7. And the standard deviation of the noise was estimated σ ∈ [1. 5,3. 5]. The peak signal to noise ratio of the image after denoising by Beyes Shrink method is improved by 3 dB when the noise standard deviationσ ∈ [1. 5,3. 5],while NLM and the algorithm proposed in this paper are useless. when σ ∈[8,15],NLM and the algorithm in this paper has equal quality,The peak signal to noise ratio of the image after denoising is improved by 7. 5 dB but using Beyes Shrink,it is only improved by 6. 5 dB; when σ ∈ [15,30],The peak signal to noise ratio of the image after denoising by the algorithm proposed in this paper is improved by10. 53 dB,which is 1. 4 times as much as the NLM algorithm and 1. 3 times as much as the Beyes Shrink algorithm. The algorithm proposed in this paper shows great superiority. The conclusion of this paper provides theoretical guidance and technical support for the accurate noise reduction of the slice images of Cleistogenes Songorica's leaf.
The microscopic images of Cleistogenes Songorica's leaf are inevitably corrupted by noise in obtaining,which will be easy to lead to the inaccuracy of its subsequent extraction and measurement of the feature parameters. Wavelet transform and curve fitting are used to determine the type and intensity of the microscopic image noise,in this paper. On this basis,wavelet threshold method,non local algorithm and the proposed combined non local mean filter( NLM) and wavelet threshold method are applied to de noise the slice images. The results show that: The noise type was recognized as additive Gaussian noise. Energy distribution of high frequency HH sub-band of 10 microscopic images randomly selected were fitted with the Gaussian function,with the R2 value of 0. 990 7. And the standard deviation of the noise was estimated σ ∈ [1. 5,3. 5]. The peak signal to noise ratio of the image after denoising by Beyes Shrink method is improved by 3 dB when the noise standard deviationσ ∈ [1. 5,3. 5],while NLM and the algorithm proposed in this paper are useless. when σ ∈[8,15],NLM and the algorithm in this paper has equal quality,The peak signal to noise ratio of the image after denoising is improved by 7. 5 dB but using Beyes Shrink,it is only improved by 6. 5 dB; when σ ∈ [15,30],The peak signal to noise ratio of the image after denoising by the algorithm proposed in this paper is improved by10. 53 dB,which is 1. 4 times as much as the NLM algorithm and 1. 3 times as much as the Beyes Shrink algorithm. The algorithm proposed in this paper shows great superiority. The conclusion of this paper provides theoretical guidance and technical support for the accurate noise reduction of the slice images of Cleistogenes Songorica's leaf.
引文
[1]张海娜,苏培玺,李善家,等.荒漠区植物光合器官解剖结构对水分利用效率的指示作用[J].生态学报,2013,33(16):4909-4918.
[2]李彦,汪胜前,邓承志.多尺度几何分析的图像去噪方法综述[J].计算机工程与应用,2011,47(34):168-173.
[3]李嘉浪,李华君,徐庆.基于小波阈值的非局部均值去噪[J].计算机工程与科学,2015,38(7):1546-1549.
[4] Buades A,Coll B,Morel J M. Nonlocal image and movie denoising[J]. International journal of computer vision,2008,76(2):123-139.
[5]翟栋,丁亚男,李涛.基于非局部的图像去噪方法的发展现状综述[J].电子科技,2014,27(1):160-162.
[6] Vignesh R,Oh B T,Kuo C C J. Fast non-local means(NLM)computation with probabilistic early termination[J]. IEEE Signal Processing Letters,2010,17(3):277-280.
[7] Van De Ville D,Kocher M. Nonlocal means with dimensionality reduction and SURE-based parameter selection[J]. IEEE Transactions on Image Processing,2011,20(9):2683-2690.
[8] Chaudhury K N,Singer A. Non-local Euclidean medians[J].IEEE Signal Processing Letters,2012,19(11):745-748.
[9] Sun Z,Chen S. Analysisi of non-local Euclidean medians and ids improvement[J]. IEEE Signal Processing Letters,2013,20(8):303-306.
[10]王海超,王春光,宗哲英,等.基于噪声类型及强度估计的狭叶锦鸡儿叶切片图像盲去噪[J].农业工程学报,2017,33(10):229-238.
[11] Tai S C,Yang S M. A design framework for hyb-rid approaches of image noise estimation and its application to noise reduction[J]. Journal of Visual Communication&Image Representation,2012,23(5):812-826.
[12] Tian J,Chen L. Image noise estimation using a variationa daptive evolutionary approach[J]. IEEE Signal Processing Letters,2012,19(7):395-398.
[13] Uss M,Vozel B,Lukin V. Image informative maps for estimating noise standard deviation and texture parameters[J].Eurasip Journal on Advance-s in Signal Processing,2011(1):1-12.
[14] Mallat S. Theory from Multi-resolution Signal Decomposition:the Wavelet Representation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1989,11(7):674-693.
[15] Liu W. Y. Image Coding and Specific VLSI Design Based on Wavelet Transform[M]. Beijing:Publishing house of electronics industry,2006:30-50.
[16] Donoho D L,Johnstone J M. Ideal spatial adaption by wavelet shrinkage[J]. Biometrika,1994,81(3):425-455.
[17]赵银善,吐尔洪江,阿布都克力木.一种改进的小波阈值函数去噪方法[J].计算机工程与应用,2013,49(22):212-214.
[2]李彦,汪胜前,邓承志.多尺度几何分析的图像去噪方法综述[J].计算机工程与应用,2011,47(34):168-173.
[3]李嘉浪,李华君,徐庆.基于小波阈值的非局部均值去噪[J].计算机工程与科学,2015,38(7):1546-1549.
[4] Buades A,Coll B,Morel J M. Nonlocal image and movie denoising[J]. International journal of computer vision,2008,76(2):123-139.
[5]翟栋,丁亚男,李涛.基于非局部的图像去噪方法的发展现状综述[J].电子科技,2014,27(1):160-162.
[6] Vignesh R,Oh B T,Kuo C C J. Fast non-local means(NLM)computation with probabilistic early termination[J]. IEEE Signal Processing Letters,2010,17(3):277-280.
[7] Van De Ville D,Kocher M. Nonlocal means with dimensionality reduction and SURE-based parameter selection[J]. IEEE Transactions on Image Processing,2011,20(9):2683-2690.
[8] Chaudhury K N,Singer A. Non-local Euclidean medians[J].IEEE Signal Processing Letters,2012,19(11):745-748.
[9] Sun Z,Chen S. Analysisi of non-local Euclidean medians and ids improvement[J]. IEEE Signal Processing Letters,2013,20(8):303-306.
[10]王海超,王春光,宗哲英,等.基于噪声类型及强度估计的狭叶锦鸡儿叶切片图像盲去噪[J].农业工程学报,2017,33(10):229-238.
[11] Tai S C,Yang S M. A design framework for hyb-rid approaches of image noise estimation and its application to noise reduction[J]. Journal of Visual Communication&Image Representation,2012,23(5):812-826.
[12] Tian J,Chen L. Image noise estimation using a variationa daptive evolutionary approach[J]. IEEE Signal Processing Letters,2012,19(7):395-398.
[13] Uss M,Vozel B,Lukin V. Image informative maps for estimating noise standard deviation and texture parameters[J].Eurasip Journal on Advance-s in Signal Processing,2011(1):1-12.
[14] Mallat S. Theory from Multi-resolution Signal Decomposition:the Wavelet Representation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,1989,11(7):674-693.
[15] Liu W. Y. Image Coding and Specific VLSI Design Based on Wavelet Transform[M]. Beijing:Publishing house of electronics industry,2006:30-50.
[16] Donoho D L,Johnstone J M. Ideal spatial adaption by wavelet shrinkage[J]. Biometrika,1994,81(3):425-455.
[17]赵银善,吐尔洪江,阿布都克力木.一种改进的小波阈值函数去噪方法[J].计算机工程与应用,2013,49(22):212-214.