Abstract: A robust bilateral filtering algorithm based on consistent subneighborhoods was presented. Firstly, employ adaptive region growing method to split local neighborhood of the seed pixel into consistent sub-neighborhoods. Then,within one of the sulrneighborhoods, smooth the value of seed pixel by using improved bilateral filter algorithm. To enhance the robust performance, by following the definition of non-local means filtering, define kernel function of this filter based on geometric closeness and local neighborhood window similarity. Because this algorithm combines both the advantages of bilateral filter and non-local means filter, and also denoises within consistent sulrneighborhoods, it can gain a more reasonable image effect. Simulation experiments demonstrate that the presented algorithm can remove noise effectively, and simultaneously preserve detailed feature of the image.

Key words: Image denoising, Bilateral filtering, Non-local means filtering, Sukrneighborhoods