A mine image denoising algorithm based on improved trimmed mean
-
摘要: 现有矿井图像去噪算法对于复杂噪声的去除效果有限，且处理速度不能满足实时监控需求。针对该问题，提出一种基于改进切尾均值的矿井图像去噪算法。首先，采用切尾均值滤波器对图像噪声进行初步滤除，同时引入二次检验机制处理残留的噪声点，通过引入离散系数提升算法对不同像素的区分能力，增强去噪性能；其次，采用基于极值数量的分类处理及再次检验机制，有效减少残留噪声问题；然后，在小波函数中引入新的控制变量优化软阈值函数和硬阈值函数，构建双阈值函数，结合Radon变换增强对线性特征的处理，增强对矿井图像的检测能力；最后，采用均方误差（MSE）与峰值信噪比（PSNR）进行图像质量评价。实验结果表明：相较于切尾均值算法、硬阈值算法、软阈值算法，基于改进切尾均值的矿井图像去噪算法处理的图像的MSE增长相对缓慢，MSE最小，图像去噪效果最好；引入离散系数后，去噪图像的MSE相较于引入前低300 dB左右，PSNR相较于引入前高20 dB左右，引入离散系数能有效减少噪声点对算法的影响；相较于卡尔曼遗传优化算法、变换域图像去噪算法、交叉分支卷积去噪网络，基于改进切尾均值的矿井图像去噪算法处理的图像MSE分别降低了27，21，13 dB，PSNR分别提升了8，6，3 dB，去噪耗时分别缩短了0.20，0.16，0.14 s。Abstract: The existing mine image denoising algorithms have limited effectiveness in removing complex noise, and their processing speed cannot meet the requirements of real-time monitoring. In order to solve the above problems, a mine image denoising algorithm based on improved trimmed mean is proposed. Firstly, a trimmed mean filter is used to preliminarily filter out image noise, and a secondary inspection mechanism is introduced to handle residual noise points. By introducing discrete coefficients, the algorithm's capability to distinguish different pixels is improved, enhancing the denoising performance. Secondly, a classification processing and retesting mechanism based on the number of extreme values is adopted to effectively reduce the problem of residual noise. Thirdly, new control variables are introduced into the wavelet function to optimize the soft threshold function and hard threshold function, and a dual threshold function is constructed. The method combines with Radon transform to enhance the processing of linear features and enhance the detection capability of mine images. Finally, mean square error (MSE) and peak signal-to-noise ratio (PSNR) are used for image quality evaluation. The experimental results show that compared to the trimmed mean algorithm, hard threshold algorithm, and soft threshold algorithm, the MSE growth of the mine image denoising algorithm based on the improved trimmed mean is relatively slow, with the smallest MSE and the best image denoising effect. After introducing the discrete coefficient, the MSE of the model is about 300 dB lower than before, and the PSNR is about 20 dB higher than before. Introducing the discrete coefficient can effectively reduce the impact of noise points on the algorithm. Compared with Kalman genetic optimization algorithm, transform domain image denoising algorithm, and cross branch convolutional denoising network, the MSE of the proposed algorithm is reduced by 27, 21, and 13 dB respectively. The PSNR is improved by 8, 6, and 3 dB respectively. The time consumption is shortened by 0.20, 0.16, and 0.14 seconds, respectively.
-
图 2 小波变换与Radon变换融合流程
Figure 2. The fusion process of wavelet transform and Radon transform
图 4 不同算法去噪图像的MSE和PSNR
Figure 4. MSE and PSNR of denoised images by different algorithms
图 5 离散系数引入前后去噪图像的MSE和PSNR对比
Figure 5. Comparison of MSE and PSNR of denoised image before and after the introduction of discrete coefficients
表 1 不同算法实验结果对比
Table 1. Comparison of experimental results of different algorithms
算法 MSE/dB PSNR/dB 去噪耗时/s 卡尔曼遗传优化算法 277 53 3.11 变换域图像去噪算法 271 55 3.07 交叉分支卷积去噪网络 263 58 3.05 本文算法 250 61 2.91 
下载: 导出CSV
-
[1] DHAS M M,SINGH N S. Image denoising using discrete wavelet transform and adaptive thresholding optimised with improved arithmetic optimisation algorithm and guided filter[J]. International Journal of Computational Intelligence Studies,2022,11(2):131-156. doi: 10.1504/IJCISTUDIES.2022.126898 [2] ZHU Yun,GONG Chengjian,LIU Shuwen,et al. Infrared object detection via patch-tensor model and image denoising based on weighted truncated Schatten-p norm minimization[J]. IET Image Processing,2023,17(6):1762-1774. doi: 10.1049/ipr2.12753 [3] WANG Yingmei,WANG Zhendong. Image denoising method based on variable exponential fractional-integer-order total variation and tight frame sparse regularization[J]. IET Image Processing,2021,15(1):101-114. doi: 10.1049/ipr2.12010 [4] DHILLON D,CHOUHAN R. Edge-preserving image denoising using noise-enhanced patch-based non-local means[J]. Multimedia Systems,2023,29(3):1025-1041. doi: 10.1007/s00530-022-01035-0 [5] REKHA H,SAMUNDISWARY P. Image denoising using fast non-local means filter and multi-thresholding with harmony search algorithm for WSN[J]. International Journal of Advanced Intelligence Paradigms,2023,24(1/2):92-109. doi: 10.1504/IJAIP.2023.128076 [6] WANG Qinglin,BAI Qing,LIU Yuting,et al. SNR enhancement for BOTDR with spatial-adaptive image denoising method[J]. Journal of Lightwave Technology:A Joint IEEE/OSA Publication,2023,41(8):2562-2571. doi: 10.1109/JLT.2022.3231973 [7] 霍一,马晓轩. 卡尔曼滤波结合遗传算法的矿井图像去噪算法研究[J]. 计算机应用与软件,2022,39(2):220-227.HUO Yi,MA Xiaoxuan. Mine image denoising algorithm based on Kalman filter and genetic algorithm[J]. Computer Applications and Software,2022,39(2):220-227. [8] 闫洪波,赵蓬勃,刘恩佐,等. 二维变分模态分解矿井监控视频图像去噪[J]. 计算机应用与软件,2023,40(6):211-215.YAN Hongbo,ZHAO Pengbo,LIU Enzuo,et al. Two-dimensional variational mode decomposition for mine monitoring video image denoising[J]. Computer Applications and Software,2023,40(6):211-215. [9] 程德强,王雨晨,寇旗旗,等. 基于改进深度残差网络的矿井图像分类[J]. 计算机应用研究,2021,38(5):1576-1580.CHENG Deqiang,WANG Yuchen,KOU Qiqi,et al. Classification of mine images based on improved deep residual network[J]. Application Research of Computers,2021,38(5):1576-1580. [10] 孙峰,李博,高紫俊,等. 一种基于Retinex理论矿井下图像增强算法[J]. 大连工业大学学报,2023,42(2):151-156.SUN Feng,LI Bo,GAO Zijun,et al. A mine image enhancement algorithm based on Retinex theory[J]. Journal of Dalian Polytechnic University,2023,42(2):151-156. [11] SINOVA B. On depth-based fuzzy trimmed means and a notion of depth specifically defined for fuzzy numbers[J]. Fuzzy Sets and Systems,2022,443:87-105. doi: 10.1016/j.fss.2021.09.008 [12] XU Yanlei,WANG Xindong,ZHAI Yuting,et al. Precise variable spraying system based on improved genetic proportional-integral-derivative control algorithm[J]. Transactions of the Institute of Measurement and Control,2021,43(14):3255-3266. doi: 10.1177/01423312211022446 [13] 文小波. M估计下切尾均值和平尾均值的抽样分布[J]. 湖南文理学院学报(自然科学版),2020,32(1):4-8,94.WEN Xiaobo. Sampling distribution of trimmed mean and winsorised mean of M estimator[J]. Journal of Hunan University of Arts and Science (Science and Technology),2020,32(1):4-8,94. [14] BORISCH E A,FROEMMING A T,GRIMM R C,et al. Model-based image reconstruction with wavelet sparsity regularization for through-plane resolution restoration in T2-weighted spin-echo prostate MRI[J]. Magnetic Resonancein Medicine:Official Journal of the Society of Magnetic Resonance in Medicine,2023,89(1):454-468. doi: 10.1002/mrm.29447 [15] SHAH S A A,BAIS A,ALASHAIKH A,et al. Discrete wavelet transform based branched deep hybrid network for environmental noise classification[J]. Computational Intelligence,2023,39(3):478-498. doi: 10.1111/coin.12577 [16] WANG Juan. Application of wavelet transform image processing technology in financial stock analysis[J]. Journal of Intelligent & Fuzzy Systems:Applications in Engineering and Technology,2021,40(2):2017-2027. [17] SINGH P,DIWAKAR M. Wavelet-based multi-focus image fusion using average method noise diffusion (AMND)[J]. Recent Advances in Computer Science and Communications,2021,14(8):2436-2448. doi: 10.2174/2666255813999200720163938 [18] WANG Siyuan,LYU Junjie,HE Zhuonan,et al. Denoising auto-encoding priors in undecimated wavelet domain for MR image reconstruction[J]. Neurocomputing,2021,437(4):325-338. [19] JIANG Yanhua,LAN Guanglin,ZHANG Zhiqing. Ship engine detection based on wavelet neural network and FPGA image scanning[J]. Alexandria Engineering Journal,2021,60(5):4287-4297. doi: 10.1016/j.aej.2021.02.028 [20] VAIYAPURI T,ALASKAR H,SBAI Z,et al. GA-based multi-objective optimization technique for medical image denoising in wavelet domain[J]. Journal of Intelligent & Fuzzy Systems:Applications in Engineering and Technology,2021,41(1):1575-1588. [21] SRIDHAR B. A wavelet based copyright marking on image under sub-bands stacking technique[J]. Latin American Applied Research,2021,51(1):71-75. [22] SOLAK A,CEYLAN R,BOZKURT M A,et al. Adrenal tumor segmentation on U-Net:a study about effect of different parameters in deep learning[J]. Vietnam Journal of Computer Science,2024,11(1):111-135. doi: 10.1142/S2196888823500161 [23] ZHAO Yan,GUO Ming,CHEN Xiangyong,et al. Attention-based CNN fusion model for emotion recognition during walking using discrete wavelet transform on EEG and inertial signals[J]. Big Data Mining and Analytics,2024,7(1):188-204. doi: 10.26599/BDMA.2023.9020018 [24] LIU Meng,MENG Kexin,XING Ruyi,et al. Haar wavelet transform and variational iteration method for fractional option pricing models[J]. Mathematical Methods in the Applied Sciences,2023,46(7):8408-8417. doi: 10.1002/mma.8343 [25] CHHA H,PENG Yongbo. Adaptive linear quadratic regulator for optimal structural control based on wavelet transform and genetic algorithm[J]. Engineering Computations:International Journal for Computer-Aided Engineering and Software,2023,40(4):1016-1039. -
点击查看大图
图(6) / 表(1)
计量
- 文章访问数: 114
- HTML全文浏览量: 37
- PDF下载量: 16
- 被引次数: 0
