Research On Image Denoising Algorithm Based On Similar Patch

Image noise is a random phenomenon caused by multiple factors.The existence of noise directly affects the expression of the real information,how to recover the original information from polluted image and improve image quality has become a hotspot of image processing research.Non local mean(NLM)is an effective denoising algorithm to remove additive Gauss noise,which has attracted the attention of many scholars for many years.NLM algorithm extends the weighted average of the similar patchs to the whole region of the image,and obtaines a better denoising effect.However,the NLM algorithm still has some shortcomings in the selection and measurement of similar patchs,parameters setting and execution efficiency.Focusing on selection of similar patchs and similarity measurement,this paper studies the improvement and optimization of NLM image denoising algorithm.The main research work includes:To address the problem of low similarity of image block in some region of NLM denoising algorithm,an improved NLM algorithm is proposed to optimize the similar patchs selection strategy.Through the superpixel segmentation algorithm,image is divided into several similar blocks.For the region with rich variation,the texture information of image blocks is used directly to choose image blocks with higher similarity.Similar windows of different sizes are adopted according to different regional characteristics.The new similar patchs selection strategy retains more texture information of original image.In order to solve the problem that the matching block with less similarity affects denoising in the recovery of the pixel information,the method of second selection for similar patchs and a new weight distribution is proposed.Considering the higher similarity between the image blocks of the same class as the center pixels,Clustering is performed in the search window according to the local features of similar patchs.Meanwhile,the image blocks in the same cluster are given higher weight.Due to the selection of a more reasonable similarity patchs,the denoising performance of the algorithm is improved.Aiming at tackling the shortage of NLM denoising algorithm that similarity is measured by using inter pixel Euclidean distance,a new decision function is used for measurement of the similarity patch.By decomposing the similar patch in multi-level wavelet,energy information is extracted from wavelet coefficients and the definition of similarity evaluation function is given based on wavelet transform.The simulation experiments show that the new similarity measurement has fulfilled good performance in image denoising,which reflects the effectiveness of the proposed method.The experimental results show that the proposed algorithms are effective and feasible.