Astronomical image restoration using an improved anisotropic diffusion

Images secured from an astronomical telescope usually suffer from blur and from interference that scientists refer to as “noise”. Therefore, good image restoration technique has become an important tool in astronomical observation. In this paper, we propose a modified anisotropic diffusion scheme to tackle the problem of image restoration in astronomy, especially in the case of nebula images. In such images, a mass of stars may be extremely bright but also may be spread randomly in dark space, and the shape of the nebula may therefore appear obscure. To restore the original appearance of a nebula, noisy stars must be filtered out and the detailed structure of the nebula must be well enhanced. The classical Perona–Malik anisotropic diffusion model that only considers gradient information cannot filter out noisy stars from the nebula image. In this study, we propose a modified anisotropic diffusion model that incorporates both gradient and gray-level variance information to remove “sparking” stars of various sizes and brightness in a nebula image. Experimental results from a number of astronomical nebula images have shown that the proposed anisotropic diffusion scheme can effectively remove noisy stars and maintain the shape of nebula in this particular case.