Search

Multi-source image fusion has been the hotspot of research in recent years, in order to make the fusion results more consistent with human vision and computer processing, many classical algorithms have been proposed, such as contourlet transform, non-subsampled contourlet transform (NSCT)and so on. On this basis, this paper proposes an improved fusion method of infrared and visible images based on NSCT transform and pulse coupled neural network (PCNN). The basic idea is that the source images are decomposed by NSCT to obtain low frequency and high frequency bands, low frequency components use the regional variance saliency fusion method of adaptive threshold weighted, high frequency components use the PCNN method, and high frequency subband coefficients are directly used as external inputs of PCNN neurons. Experimental results show that the fused image has higher clarity and richer useful information. Both subjective and objective performance indicators are better than the traditional algorithm.

In order to improve the quality of fusion image integrated by infrared and visible images, so that it is more suitable for human and computer vision or processing. A two-step fusion method, that is region structure similarity and Contourlet transform, is presented. With the blocking idea, source images are fused by using region structure similarity, region energy ratio and region clarity ratio to obtain the first fusion image. Then the first fusion image and two source images are decomposed by taking Contourlet transform, and low frequency components use region variance weighting rule, high frequency components adopt multi-scale analysis. Experimental results prove that our proposed method can improve subjective visual effect of the image. Meanwhile objective evaluation indexes entropy, standard deviation, average gradient and mutual information are increased.