Multi-Focus Raw Bayer Pattern Image Fusion for Single-Chip Camera

摘要

In this paper, an efficient patch-based image fusion approach for rawrnimages of single-chip imaging devices incorporated with the Bayer CFA pattern isrnpresented. The multi-source raw Bayer pattern images are firstly parted into halfrnoverlapped patches. Then, the patches with maximum clarity measurement definedrnfor raw Bayer pattern image are selected as the fused patches. Next, all the fusedrnlocal patches are merged with weighted average method in order to reduce thernblockness effect of fused raw Bayer pattern image. Finally, the real color fusedrnimage is obtained by gradient based demosaicing technology. The multi-source rawrnBayer pattern data is fused before demosaicing, so that the multi-sensor system willrnbe more efficient and the artifacts introduced in demosaicing processing do notrnaccumulate in image fusion processing. For comparison, the raw images are alsorninterpolated firstly, and then various image fusion methods are used to get the fusedrncolor images. Experimental results show that the proposed algorithm is valid andrnvery effective.