Research On Wavelet Image Denoising Algorithm Based On Improved Threshold Function

Image denoising is a critical step in image processing and has always been the research focus of scholars.The question of image denoising is how to better eliminate the noise information in the image while retaining the real signal of the image.In recent years,wavelet threshold and nonlocal mean denoising algorithms have attracted a lot of attention because of their excellent denoising performance.Therefore,taking these two ways as the object of study,the research of this paper includes the under contents:(1)A wavelet denoising algorithm based on CALE threshold function is proposed.Firstly,proposes a new threshold.The new threshold will select a threshold that is more consistent with the noise figure with the change of the number of decomposition layers.Then a continuous and low error threshold function is designed.The threshold function has better continuity at the threshold.When the threshold tends to infinitely great,the difference between the estimated coefficient and the real coefficient also tends to zero.Therefore,the image detail information after denoising is saved more completely.(2)An improved MNLM image denoising algorithm is proposed.Aiming at the problem that the traditional block level NLM image denoising algorithm has the problem that the image blocks with high similarity can calculate the smaller weight value,firstly,decomposes the image at the block level multi-scale,and then designs an improved weight function with faster attenuation speed,The weight function can better avoid the problem of non-zero weight between different image blocks.Make the weight value better match the image block with high similarity,The estimated pixel value obtained after denoising is closer to the real pixel value.Lastly the improved MNLM image denoising algorithm is applied to the mixed noise removal of fingerprint image.Experimental proof that the fingerprint image denoised by this algorithm has good definition.(3)In order to verify the effect of the two improved algorithms,lena image and fingerprint image are used as test images.Gaussian noise with different density is added to lena test image,and gaussian and salt and pepper mixed noise with different density are added to fingerprint test image,experiments were carried out separately.The simulation results show that the denoising effect of the two algorithms is better than that of other comparison algorithms,and more complete image details are retained after denoising,and significantly improve the image quality.Figure [23] Table[8] Reference[69]...