Infrared image mixed noise removal method based on improved NL-means

Typical denoising algorithm were unable to effectively remove the mixed noise of stripe and random noise in infrared images. To solve this problem, an improved NL-means filter was proposed. Firstly, the problem in NL-means algorithm dealing with mixed noise was analyzed, and an experiment was performed to analyze the characteristic of the mixed noise in image patches of infrared images. Based on the experiment results, the mixed noise was modelled using colored Gaussian model, and the ordinary patch similarity index was improved using Mahalanobis distance instead of Euclidean distance, so that it adapts to the local with different complexity in the image. Both simulated and real data experiments show that the algorithm can effectively remove the stripe and random mixed noise compared with traditional image denoising algorithms.