Research Of Fusion Method For Multifocus Images

Multifocus image fusion technology is a major study in the field of image fusion. With the method of image processing, this technology can effectively solve the limit of focus range on optical lens. Therefore,objects of different distances in the environment can display more clearly in the same image. At the same time,this technology could acquire superior foundation for the follow-up processing,such as feature extraction,edge detection and target recognition. Furthermore, this technology can remove the redundant information and improve information utilization in the image. Multifocus image fusion technology play an important role in scientific research, daily life and other aspects.In fusion methods for multifocus images, the methods based on space domain can directly implement image fusion in the pixel space, which can fully retain the original pixels in images. In addition, higher fusion speed and superior local pixels consistency could also be acquired by this method. However, because of the complexity of the image content. It's difficult to get the focus areas information in the source image into fusion image accurately and fully,such as the blocking-artifacts of fusion images in the block-based methods,the spatial misalignment of fusion images in the region-based methods, etc.This paper focus on fusion methods based on spatial domain for multifocus images. In view of the potential defects and deficiencies of existing image fusion methods, the corresponding solutions is proposed. The main contribution of this paper can be concluded as follows.(1)An intelligent block-based fusion method is is presented. In order to reduce the blocking-artifacts of fused images. An evaluation index for fusion image quality named LUE-SSIM was firstly proposed, which utilizes the characteristics of human visual system and structural similarity to make the index consistent with the human visual perception. Particle swarm optimization(PSO) algorithm is used for optimizing the block size to construct the fused image. Experimental results on LIVE image database shows that LUE-SSIM outperform SSIM on Gauss blur images quality assessment. Besides, the results of 12 groups multifocus image fusion experiments show that the proposed method performs better than other block-based methods, especially in aspect of reducing the blocking-artifact of the fused image. And find that by emphasizing the expression of undistorted edge information can effectively reduce or weaken the blocking-artifact of fusion images.(2) A fusion method based on DT-CWT and focused region segmentation is presented. This method can implement the image fusion with the focused area, which has many remarkable advantages. In order to reduce the spatial misalignment of fusion images, we utilize dual tree complex wavelet transform to realize the focus area segmentation.And do some subsequent image processing to improve the accuracy and completeness of region segmentation. As for the design of fusion scheme, the source image was divided into three areas to realize the smooth connection between different areas. Compared with different fusion rules in DT-CWT, the proposed rules have higher sensitivity in contrast information detection, and can enhance the consistency of local pixels. Multi-focus image fusion results on 12 groups images show that this method can effectively increase the accuracy of focused area segmentation, and reduce the spatial misalignment of fusion images.The multifocus image fusion methods in this paper turn to be stable and reliable, which can enrich the current theories of multifocus image fusion and the future research on this area. Using those methods can have a positive impact on Image fusion applications.