Research On Fusion Algorithm Of Infrared And Visible Light Images Based On Multi-scale Decomposition

With the industry’s development trend and the advantages of image categories,image fusion research based on Infrared(IR)and Visible Light(VIS)has become a hot research topic and focus in the field of information fusion assignment.The key to fusion is t he effective integration of important targets in IR images and rich details in VIS images,which is important for highlighting targets,enhance understanding of effective information in images,and subsequent applications.The Multi-scale Transform(MST)Methods multi-resolution,multi-direction,and anisotropy can effectively display the image’s information features.As a consequence,the multi-scale decomposition and fusion of IR and VIS images based on the Non-subsampled Contourlet Transform(NSCT)metho d in MST has low contrast,insignificant objectives,loss of texture details,and the introduction of edge artifacts,and other common problems in to carry out improvement research in order to create more effective and adaptable algorithms.The following ar e the major previous research:(1)The traditional fusion method based on pixel and window fusion rules and the single feature fusion target information are not prominent enough,and the details and textures are seriously missing.Based on multijudgment and weighted least squares optimization,this paper proposes a fusion method for NSCT infrared and visible image fusion.To continue,the multi-scale decomposition process obtains the image’s low-frequency and high-frequency subbands.Second,the low-frequency subbands local squared entropy and sum-modified Laplacian(SML)are selected to complement each other,with aim of separating a small amount of detail information under good contrast.The high-frequency subbands fully consider the importance of the underlying features and selects Phase Consistency(PC),Weighted Sum-modified Laplacian(WSML),and Weighted Local Energy(WLE)fusion,and then performs Weighted Least Squares(WLS)optimization on it,which can extract more edge texture information and red uce irrelevant details such as noise.The experimental results show that the method proposed in this paper is better in object saliency,image contrast,and information extraction.At the same time,while ensuring that the obj ective evaluation index Mutual Information(MI)has a good result,the Average Gradient(AG),Information Entropy(IE),Spatial Frequency(SF),and is an optimal result.Explain that the proposed algorithm makes up for the deficien cy of single feature extraction,and is an effective method to enhance image quality.(2)The traditional and improved Saliency Detection algorithms based on regional fusion rules for infrared images are sensitive to noise,are simple and effective only for a few simple images,and have poor adaptability and anti-interference ability.An IR and VIS image fusion method model for NSCT based on improved frequency-tuned(FT)Saliency Detection is proposed.To extract a better IR saliency map,the limited grayscale range,Gaussian filtering replaced by guided filtering,grayscale energy,and contrast stretching function enhancement processing aiming at infrared images are used to distinguish the target from the background.For the low-frequency subbands coefficients,the calculated infrared saliency weight map is used to guide the fusion,and for the high-frequency part,the rules of weighted local energy and WLS optimization are used to extract the detailed-rich images.The experiment first adopts the average value of AG,IE,SF,and MI as a fast evaluation method,which verifies the ability of the saliency detection algorithm in this paper and determines the optimization of the combined parameters.Then select four groups of images in the database and five traditional algorithms for qualitative and quantitative analysis.The subje ctive judgment and the best results of the four indicators show that the fusion method in this paper is more prominent in the ability to extract visible detail information and retain the target energy,and it also shows a good ability to suppress edge artifacts.It is an improved,effective and robust method.