A denoising method is proposed for full body X-ray images, acquired under low dose conditions. The suggested algorithm is based on a non local means filter adapted to the statistics of Poisson noise. A new feature of the method is to locally set the filtering parameters in order to denoise while preserving details in low absorption regions. Thus, we propose to adapt the denoising parameters for each pixel by exploiting a global noise level measure and the standard deviation image of the gradient magnitude. Quantitative and visual results on phantom and real images show the interest of the method, achieving the objectives.