- In this work we suggest a novel model for automatic noise estimation and image denoising. In particular, we investigate the useful affinity which arises between statistical mechanics and image processing, and describe a framework from which novel denoising algorithms can be derived: Ising-like models and simulated annealing techniques. This is the first time such algorithms are used for colored images and video denoising. Results, as well as benchmarks, suggest a significant gain in PSNR and SSIM in comparison to other filters, mainly in cases of low impulse noise. When hybridizing our models with other image processing techniques they are shown to be even more effective. Their major disadvantages- high complexity and limited applicability, are also discussed. We present a detailed analogy between image processing and statistical physics.We use a novel Ising-like model for denoising of damaged colored images and videos.The algorithm chooses automatically the denoising parameters.A novel simulated annealing code, based on the L1 norm is then run.We denoise images and videos damaged by additive impulse or Gaussian noise.The restoration results are better compared to other well-known filters.