An Image Denoising Algorithm Based on Improved Median Filter and Threshold Function

Mixed noise is commonly present in images, posing challenges for traditional denoising algorithms such as median filtering, mean filtering, and wavelet threshold filtering. In this paper, a novel approach is proposed for denoising mixed noise. The proposed method combines an improved median filtering technique with a wavelet thresholding function. By incorporating the strengths of both methods, our algorithm aims to effectively remove mixed noise from images. Firstly, we enhance the performance of the median filtering algorithm to provide initial denoising. Subsequently, a two-level wavelet transform is applied to the pre-filtered image, enabling the extraction of high-frequency components. These components are then filtered using an improved thresholding function and median algorithm. Finally, an inverse wavelet transform is performed to reconstruct the denoised image. The experimental results show that our algorithm improves the peak signal-to-noise ratio (PSNR) of the denoised image by 2.17%~10.56%, and reduces the root-mean-square error (RMSE) by 6.24%~25.19% compared to the algorithms such as the median filter, wavelet soft-threshold filter, wavelet hard-threshold filter, and wavelet semi-soft threshold filter, etc. Hence, the proposed algorithm in this paper exhibits superior preservation of image details and more efficient removal of mixed noise.