Research On Infrared And Visible Image Fusion Technology Based On NSST

In the recent years,Non-subsampled Shearlet Transform(NSST)as a continuously growing image fusion technique has been widely used in field of infrared and visible light image fusion due to multi-scale and multi-directionality,low time complexity.In the thesis,the advantages of NSST transform in the field of fusion and the characteristics of infrared and visible light images are fully taken into consideration.And the research of infrared and visible light image fusion methods is carried out under the NSST decomposition framework,in order to obtain more high-quality fusion images.And the significant research work of the thesis is shown as follows.1)To improve or even solve the problem of low contrast and loss of background detail information retention in infrared and visible image fusion methods,a method combining adaptive pulse coupled neural network(PCNN)with NSST is proposed.NSST is used to decompose the source image into low-frequency sub-bands and high-frequency sub-bands at multiple scales;Next,for the fusion of image high and low frequency subbands,the fusion rules of weighted average and adaptive PCNN are used respectively;Finally,the high-frequency images and low-frequency images are processed by NSST inverse transformation to form a fusion image.The simulation results indicate that the method proposed selected,the contrast of the fused image is improved,the background details is preserved and a better fusion effect is achieved.2)The infrared and visible image fusion methods have problems of unsatisfactory edge detail information retention.To avoid this situation,a fusion method combining NSST and guided filtering is considered.The adaptive fuzzy logic method is used to perform low-frequency sub-band fusion,and the method of regional energy maximization is applied to perform high-frequency sub-band fusion.According to the experiment,it is showed that with the method used,the contrast of the fusion image is improved,the edge information of the fusion image is retained,and it makes the fusion image more conducive to human observation.3)The infrared and visible image fusion method that is based on the combination of convolutional neural network(CNN)and NSST is considered to preserve the salient features of the source image as much as possible,and solve the problem that the feature is difficult to extract.The CNN and frequency modulation algorithm is used to extract the saliency of the target.Next,the image fusion is achieved with target saliency and NSST combined.It is indicated that with the method that is proposed chosen,the salient features of the source image is retained,the contrast of the fused image is improved,and is more conducive to direct observation by the human eye.In summary,with the method proposed in the thesis,the quality of the fusion image is improved and the fusion effect.is enhanced.At the same time,compared with traditional image fusion algorithms,this method improves objective evaluation indicators such as information entropy,average gradient,spatial frequency,mutual information,and cross entropy.