時頻濾波器過濾影像雜訊之效果

Time-Frequency Filters and Their Results on Image Noise Filtering

紀凱屏(Kay-Ping Chi) ； 蔡展榮(Jaan-Rong Tsay)

《航測及遙測學刊》 7卷4期 (2002/12) Pp. 1-16

https://doi.org/10.6574/JPRS.2002.7(4).1

摘要

本文探討各種不同的影像雜訊過濾方法，包括空間域的方法、頻率域的方法、時頻域的方法。並以影像復原概念為基礎，將五張不同類型的實驗影像分別加入隨機雜訊（如：高斯雜訊、斑駁雜訊、椒鹽雜訊）、系統性雜訊（如：相干雜訊），文末再以時頻域中的Daubechies小波法過濾衛載合成孔徑雷達影像為例，綜合分析各法過濾影像雜訊的效果。由實驗結果可看出應用時頻域的方法過濾雜訊確實是可行的，其中，空間濾波器與Daubechies小波法皆能有效的過濾高斯雜訊與斑駁雜訊；椒鹽雜訊則以空間濾渡器的中值濾渡器效果最好；而相干雜訊的部分以頻率濾波器中的帶阻濾波器效果最好。

關鍵字

時頻濾波器 ； 影像雜訊 ； 影像復原 ； 小波轉換

並列摘要

In this paper, we discuss different approaches for reducing image noise, including filtering image noise in spatial domain, frequency domain, and time-frequency domain, and estimate their denoising effect. We use a spaceborne SAR (=Synthetic Aperture Radar) image and add several kinds of noise in five different test images under the theoretical concepts of image restoration. They include (1) random noise such as Gaussian noise, speckle noise and salt-and-pepper noise, and (2) systematic noise such as coherent noise. The results demonstrate that Daubechies wavelet approach is efficient. Spatial filters and Daubechies wavelet approach can filter out the Gaussian and speckle noise well. Median filters are good ones for reducing salt-and-pepper noise. Bandreject filters are suitable for filtering out the coherent noise.

並列關鍵字

time-frequency filter ； image noise ； image restoration ； wavelet transform