Multi-Source Image Fusion Using Parameter Optimized PCNN in LNSST Domain

Abstract

A fusion method combining LNSST and PCNN is proposed to address the problem of unstable operation of LNSST and PCNN, which leads to loss of details and poor visual perception of fused images. Firstly, a division fruit flying optimization algorithm is proposed to optimize the parameters of PCNN to eliminate the effect of statistical differences in the source images on the PCNN operation. Then, the source image is decomposed into high and low frequency subbands by LNSST, and the low frequency subbands of LNSST are stimulated with a bootstrap filter to show more detail features. Finally, the high and low-frequency subbands of the LNSST are fused by the parameter optimized PCNN to complete the subband fusion; the fused image is obtained after inverse LNSST. Experimental results on 90 sets of images in three types of fusion tasks, namely multi-focus, infrared and medical, show that the method has advantages in terms of subjective image sharpness compared with other fusion methods, and objective improvement of 2.9%~29.3% in six evaluation indexes, respectively. The proposed method effectively preserves detail and texture information in different fusion tasks and improves fusion accuracy.