Image restoration is a keen problem of low level vision. In this paper, we propose a novel - assumption-free on the noise model - technique based on random walks for image enhancement. Our method explores multiple neighbors sets (or hypotheses) that can be used for pixel denoising, through a particle filtering approach. This technique associates weights for each hypotheses according to its relevance and its contribution in the denoising process. Towards accounting for the image structure, we introduce perturbations based on local statistical properties of the image. In other words, particle evolution are controlled by the image structure leading to a filtering window adapted to the-image content. Promising experimental results demonstrate the potential of such an approach.