Self-Adaptive Non-Local Means Image Denoising Algorithm Based on Fuzzy Edge Complement

In view of the holding unsatisfactory effects of the traditional non-local means algorithm for texture details, an improved non-local means denoising algorithm combined with fuzzy edge complement (FEC) is proposed. The edge texture feature image is detected by the FEC algorithm. The similarity weight parameter is chosen adaptively according to the edge texture feature, and the edge texture region and the flat region are smoothed for different degrees pertinently, which prevents the edge texture information from being lost. The similarity weights of non-local image blocks are improved using the structural similarity of edges. The effects of pixels in the same area are increased, and those in different areas are reduced. Thus, the better texture hold effects can be achieved. Experimental result indicates that the image denosing can be effectively achieved by this method. Meanwhile, the more texture detail features and geometrical structural features are persevered.