Improved Anscombe transformation and total variation for denoising of lowlight infrared images

The infrared image is corrupted by heavy Poisson noise in lowlight situation, which results in the loss of detail information. To suppress the heavy Poisson noise and keep the distinct edges of images in the lowlight situation, the denoising method based on the improved Anscombe transformation in wavelet domain is proposed. The Anscombe transformation in wavelet domain is improved to control the image distribution from Poisson into Gaussian distribution accurately; then the improved total variation regularization is applied to the wavelet-Anscombe domain for suppressing the heavy noise effectively with the optimal wavelet function. Denoising experiments on artificially degraded and practical lowlight infrared images show that the proposed denoising method can suppress the noise effectively and preserve the detail of images with heavy noise compared with several state-of-the-art methods.