Research On The Infrared And Visual Images Fusion Algorithm Based On The Non-subsampled Shearlet Transform

Image fusion is a new technology that integrates many sciences,such as sensor technology,image processing,signal processing,computer engineering and artificial intelligent.Infrared and visible image fusion is one of the hot topics in this field and plays an increasingly significant role in enhancing the performance of space monitoring,military and security monitoring.However,the current fusion results are faced with many problems,such as false information and noise,low utilization of fusion information and insufficient image edge preservation.To solve these problems,the Non-subsampled Shearlet Transform(NSST),which is an advanced multi-scale analysis tool,is introduced in this dissertation.Some researches of infrared and visible image fusion algorithm based on the NSST are conducted.The main results and innovation of this dissertation are summarized as the following aspects.1.Aimed at the problem of the existing fusion methods of visible and infrared images based on the sparse representation suffering from the noise,a fusion method of infrared and visible image based on dictionary learning and NSST is proposed.After the NSST decomposition,block vectorization,zero-averaging,sparse decomposition and a regional energy fusion rules are used in the low frequency subband.And a Laplacian energy sum fusion rule is choosed in the high frequency subbands.The simulation results show that the proposed algorithm is superior to other fusion algorithms in visual and objective evaluation.2.Pulse Coupled Neural Networks(PCNN)has a strong biological background and the style of signal processing,which is more in line with the physiological mechanism of the human visual system.PCNN has some advantages than traditional methods,however it cost a long operation time and easy affected by the noise.So a region dual-channel unit-linking PCNN(RDU-PCNN)was constructed which is more in accord with the human sense system than PCNN and own some excellent characters,such as regional characters and global coupling.Based on this,an image fusion algorithm based on RDU-PCNN and independent component analysis(ICA)bases in non-subsampled Shearlet transform(NSST)domain for infrared and visible images is proposed.Firstly,NSST is employed to decompose the source images into a series of high frequency sub-images and one low frequency sub-image.Secondly,a rule based on RDU-PCNN was put into high frequency sub-images and a rule based on ICA bases was input into low frequency sub-image.Finally the fusion image was reconstructed by an inverse NSST on these merged coefficients.Experimental results demonstrate that the proposed method can significantly improve the fusion quality.Another image fusion algorithm based on RDU-PCNN and DCT is presented.After NSST decomposition,a fusion rule based on RDU-PCNN is used in the low frequency subbands and a fusion rule based on DCT is used in the high frequency subbands.The fusion image can be obtained by taking an inverse NSST.Experimental results show that our proposed algorithm can enhance the feature in infrared image and is superior to other fusion algorithms in visual and objective evaluation,when compared with other currently popular or more advanced methods.3.Aimed at the problems that most existing fusion methods suffering from one or more imperfections such as the details information lost,too much redundant information in high frequency after NSST decomposition,an effective fusion method based on online same scene independent component analysis bases(OSS-ICA-bases)and compressed sensing(CS)is proposed.Firstly,NSST is employed to decompose the visible and infrared images into high frequency subbands and low frequency subbands.Secondly,a fusion algorithms based on CS was put into high frequency subbands and a fusion algorithms based on OSS-ICA-bases was input into low frequency subbands.Finally the fusion image was reconstructed by an inverse NSST on these merged coefficients.Because OSS-ICA-bases can suppress the noise and fuses the complementary information well,CS enables the high frequency subbands to be accurately reconstructed from fewer sparse fused coefficients,NSST can obtain the asymptotic optimal representation and has the better sparse representation ability,the proposed algorithm should obtain a better result.Experimental result also show that our approach can achieve better performance than other methods in terms of subjective visual effect and objective assessment4.Aimed at the problem for target edge in infrared image which contains noise and dim target edge in the background which is difficult to extract and artificial side effects in the fused image which leads to blurred edges and artifacts appear by traditional image fusion methods based on multi-resolution analysis,A novel and effective image fusion method based on guided filtering and phase congruency in NSST domain is proposed.Firstly,NSST is employed to decompose the visible and infrared images into detail layers and one base layer.Then,phase congruency is adopted to extract the saliency maps from the detail layers and a guided filtering is designed to compute the filtering output of base layer and saliency maps.Next,a novel weighted average technique is used to make full use of scene consistency for fusion and obtaining coefficients map.Finally the fusion image was acquired by taking inverse NSST of the fused coefficients map.Experimental results show that theproposed method get a better performance comparable with other fusion approaches of visible and infrared images.