A successful approach for denoising, especially in medical applications, is Non-local means (NLM) filtering. Recently several variations of the basic NLM algorithm have been proposed. In this work we propose a modified NLM approach. NLM approach exploit redundancy in images. We have considered the basic NLM algorithm, where, in order to modify a pixel, the values of all the pixels in the image are considered. The modified pixel value is a weighted average of all pixels, where the weight is determined based on the similarity between the neighborhood of the pixel being modified and the current pixel. In the proposed work we replaced the L2 norm used in the basic NLM algorithm with L3 norm. Experiments carried out with a set of natural and medical images shows improvement in the quality of the reconstructed images, in terms of PSNR as well as SSIM. On the negative side, there is a slight increase in the execution time.