Application of wavelet transform-based Wiener filtering method to denoise in agricultural product images

Denoising is one of the most basic and the most important task in the agricultural product image processing. In order to effectively denoise in the agricultural products image, enlightened from two-dimensional discrete Wiener Filter arithmetic, the authors put forward a wavelet transform-based Wiener Filtering method. The method combined wavelet transform with Wiener filtering, and had the advantages of sparseness, multi-resolution, getting rid of pertinence, flexibility in choosing basis, and optimum estimation of image from the meaning of MSE. First, agriculture noise image "ano" was processed by wavelet transform to have a low frequency image "a1" and three high frequency images "hd1", "vd1" and "dd1" from horizontal, vertical and diagonal directions; Second, low frequency image al was processed by Wiener Filtering to have a image "a1w", then three high frequency images were processed separately by Wiener Filtering and compounded to a image "g1w"; Third, low frequency image "a1w" and high frequency image "g1w" were transformed conversely by wavelet to a filtering image "a1w+g1w". Simultaneously, considering noise presents to high frequency mostly, so low frequency image al and high frequency image "g1w" were transformed conversely by wavelet to a filtering image "a1+g1w". This was the instance of wavelet transform to noise image for the first time, and the second, third and fourth transform were resemblant. In this way, people can have many filtering images, then finally made sure the best denoised agricultural product images according to image signal-to-noise(PSNR) and visual effect. The method was applied in agricultural product image denoising such as Chinese date and wheat weed, as a result PSNR was 158.23(visual effect was clear), and better than other methods such as neighborhood average(PSNR was 154.14), median filter(PSNR was 155.82), mathematical morphology(PSNR was 154.07, visual effect was a bit black), Gauss filter(PSNR was 153.79, visual effect was very black), direct Wiener filter(PSNR was 154.14) and wavelet denoise(PSNR was 158.18) etc. The experimental results show that wavelet transform-based Wiener filtering method applied in agricultural products image denoising has the advantages of high signal-to-noise, good visual effect; so wavelet transform-based Wiener filtering method applied in agricultural product image denoising is effective and practicable.