Integral non-local means algorithm for image noise suppression

The non-local means (NLM) algorithm computes the weighted average of the difference between two local patches centred at each neighbour and the centred pixel, which is adapted for the adaptive mean filter of noise suppression. However, computation of the weighted average of the local patch difference is very time consuming. To solve this problem, a fast weighted average computation using an integral image for the NLM, called integral NLM (INLM), is proposed. The performance of the INLM algorithm is estimated and the experimental results show that the outcomes of INLM are totally the same as those by the classical NLM, while the running time is reduced significantly.