An adaptive denoising method for tongue images in traditional Chinese medicine

This paper presents an adaptive method for removing noises of tongue images, which combines the wavelet-transform-denoising method and median-filter-denoising method. First, a new adaptive median filter which detects the edge and impulse of the image and then adaptively changes the size of the window based on the “noise-free pixel” and “noise pixel” marked previously, was used to filter the salt and pepper noise in the original image. Second, wavelet soft threshold denoising was used to remove the rest noise of the processed image in the first stage. The soft thresholding is based on Birge Massart strategy to set the value of threshold for each layer of wavelet decomposition. SNR and RMSE were utilized to select wavelet function and wavelet decomposition scale. The experiments show that the proposed method can effectively reduce the noises and achieve the highest SNR and lowest RMSE, and at the same time, preserves the details of the edges of the tongue image.