Image De-Noising Based on Association-Prediction Model

In contrast to existing image de-noising algorithms with the problems of threshold selection, parameter initialization, and model complexity, the method proposed in this paper is modeled on the fast and efficient characteristics of the optic nerve of human optic de-noising systems. Based on the association-prediction model, the image de-noising method abstracts and simulates the structure of the human brain nerve. It imports de-noising prior knowledge that is useful when the de-noising process separates the image into patches. Unlike generic image de-noising methods, the proposed method can preserve the high frequency characteristics of the image. Furthermore, it does not require the noise intensity or noise models to be estimated, thus, errors caused by noise estimation are avoided. Experiments were carried out on the simulated noisy images and real noisy images from surveillance cameras, and the results suggest that the proposed method outperforms traditional methods, proving its effectiveness and practicality.