Contourlet变换为农产品图像去噪的有效方法

农产品图像去噪是农业图像处理中最基本、最重要的工作之一。现有小波去噪方法存在各向同性的缺陷,从而限制了其去噪的效果。针对这一问题,提出了一种基于Contourlet变换的农产品图像去噪算法,该方法充分利用了Contourlet变换具有的多分辨率、各向异性和稀疏性的优点。算法首先对含噪农产品图像进行塔形方向滤波器组(pyramidal directional filter bank,PDFB)分解,然后通过多尺度萎缩阈值进行高频子带去噪,保留信号系数并抑制噪声系数,最后通过Contourlet反变换得到去噪后的图像,实现农产品图像的去噪。为了验证Contourlet变换的去噪效果,分别采用小波去噪、中值滤波、均值滤波、高斯滤波和维纳滤波对常见农产品图像进行了对比试验。试验结果表明,基于Contourlet变换的图像去噪方法应用于农产品图像去噪具有信噪比高、视觉效果好等优点,将Contourlet变换用于农产品图像去噪是有效的、可行的。

Contourlet transform as an effective method for agricultural product image denoising

Image denoising for agricultural product image is a one of the most basic and important step in agricultural image processing. Wavelet transform has the weakness of isotropy, which limits its use in image denoising. To solve this problem, a new image denoising algorithm based on Contourlet transform is presented. The algorithm fully utilized the advantages of Contourlet transform such as flexible multi-resolution, anisotropy and a sparse representation. In the first step, the image is decomposed by PDFB (pyramidal directional filter bank), and in the second step, the muti-scale threshold shrinkage algorithm is presented to remove the noise in high frequency sub-band, in the last step, inverse transformation of Contourlet is used and the agricultural product image denoising is realized. In order to test the performance of Contourlet denoising algorithm, a comparative test is made by using Wavelet, median filter, mean filter, Gaussian Filter and Wiener filtering methods. Results show that Contourlet denoising algorithm is suitable for agricultural product images and it also has the advantage of PSNR (higher peak signal to noise ratio) and visual effect. The algorithm proposed is practical and valid for agricultural product image denosing.