Multifocus image fusion via fixed window technique of multiscale images and non-local means filtering

In multifocus image fusion, accurate detection of the focused pixel from source images is crucial to improve the quality of fused result. Traditionally, the method developed in spatial domain is commonly used. However, such approaches tend to produce boundary seams or distortions. To this end, we propose a new fixed window technique of multiscale image analysis (MIA) and a new weighted fusion strategy by employing non-local means filtering (NLF). This new scheme consists of three parts: detection of focused pixel, correction of detecting results, and generation of fusion weight maps for source images. To improve detection robustness against the size of object, we develop the fixed window technique of MIA to detect the focused pixels, and then we construct the initial fusion decision map for each of source images by combining those detection results; second, we present a new refining process based on block consistency evaluation for correcting the initial detection result. At last, the corresponding source images are used as a guide and combined with the NLF to produce the fusion weight maps. Experimental results demonstrate that the performance of our approach is superior to that of many state-of-the-art algorithms in terms of both visual perception and objective evaluation.