陳家益 董夢藝 戰蔭偉 曹會英 熊剛強



摘 ?要: 對于高密度的脈沖噪聲,現有濾波算法的去噪性能并不理想,在噪聲檢測與噪聲濾除上存在缺陷。鑒于此,提出中值檢測的迭代中值濾波算法,對噪聲檢測和噪聲濾除的方法分別進行有效的改進。算法用灰度最值進行噪聲檢測,再用鄰域中值作進一步的檢測。對于噪聲像素,運用迭代的方法,用鄰域中信號像素的中值取代,充分利用了前次濾波的結果。實驗結果證明,相對于現有的濾波算法,所提出的算法有著更好的濾波性能,在濾除噪聲的同時,很好地保持了圖像的紋理邊緣和細節。
關鍵詞: 圖像去噪; 噪聲檢測; 噪聲濾除; 迭代中值濾波; 加權中值濾波; 中值檢測
中圖分類號: TN911.73?34; TP391 ? ? ? ? ? ? ? ? ? ?文獻標識碼: A ? ? ? ? ? ? ? ? 文章編號: 1004?373X(2020)07?0070?04
Iterative median filtering algorithm based on median detection
CHEN Jiayi1, DONG Mengyi2, ZHAN Yinwei3, CAO Huiying1, XIONG Gangqiang1
(1. School of Information Engineering, Guangdong Medical University, Zhanjiang 524023, China;
2. Second Clinical Medical College, Southern Medical University, Guangzhou 510515, China;
3. School of Computer Science and Technology, Guangdong University of Technology, Guangzhou 510006, China)
Abstract: The denoising performances of the existing filtering algorithms are undesirable for removing the high density impulse noise, and they have defects in noise detection and removal, so an iterative median filtering based on median detection is proposed to effectively improve the techniques of noise dection and removal. In this algorithm, the gray extreme intensity value is adopted to perform noise detection, and then the median of neighborhood is used to perform further noise detection. The signal pixels in neighborhood are used to replace the noise pixels by iterative method, in which the previous denoising result is fully taken. The experimental results show that, in comparison with the existing filtering algorithms, the proposed filtering algorithm has better filtering performance and can perfectly maintain the texture edges and details of image while removing the noises.
Keywords: image denoising; noise detection; noise filtration; iterative median filtering; weighted median filtering; median detection
0 ?引 ?言
圖像在拍攝和處理的過程中,經常受到噪聲的破壞,噪聲會影響圖像的視覺效果和圖像的處理與分析,去噪非常必要。脈沖噪聲是最常見的一種噪聲,隨機地均勻分布于圖像中,脈沖噪聲分為隨機值脈沖噪聲和固定值脈沖噪聲。
隨機值脈沖噪聲隨機地將圖像的像素灰度改為介于最小灰度值與最大灰度值之間的隨機值;固定值脈沖噪聲隨機地將圖像的像素灰度改為最小灰度值或最大灰度值[1]。對于固定值脈沖噪聲,均值濾波算法[2?4]會破壞圖像的細節和邊緣,產生模糊效果;中值濾波算法[5]屬于非線性濾波,因其具有良好的去噪性能而被廣泛應用于脈沖噪聲的去除。
標準的中值濾波算法對噪聲圖像的所有像素,統一用鄰域像素的中值替代,去噪處理具有盲目性,破壞了部分像素的原信息?!?br>