Research On Image Fusion Method Based On Thermal Radiation Information

As the single information provided by common single sensor can no longer meet the needs of people’s life,and the information obtained by multi-source sensors and fuse them according to the corresponding rules can be used to obtain more comprehensive information and make up for the defects of single sensor.Infrared images can show the temperature distribution and thermal radiation information of objects,visible images can show the shape and color information of objects,fusing the two together can obtain clearer and more accurate images,helping to improve the performance of target detection,identification and tracking,the field has been rapidly developed in recent years in China,the fusion of the two images can complement each other’s shortcomings,usually used in night vision,security monitoring,the medical diagnosis and other fields.The multiscale transform has gained widespread attention because it can decompose the image flexibly and extract the features of the image effectively.Based on the existing multiscale transform algorithms,deep learning of neural networks and adaptive fusion algorithms,the fusion process focuses too much on the salient target of the image and improves the image contrast,while ignoring the important infrared thermal radiation information,which cannot better retain the target information of infrared images,the following research has been carried out:1.The image fusion method of multi-scale transform is analyzed and studied,and the Laplace pyramid transform,wavelet transform,and non-downsampled contour wave transform are studied in detail,and the performance of each algorithm is analyzed by doing comparative experiments.2.To address the situation that current image fusion algorithms focus too much on highlighting the important targets of the image,but ignore the important infrared thermal radiation information,a fusion rule based on thermal radiation information preservation is proposed,which can preserve the thermal radiation information of the infrared image to the maximum extent,the low-frequency part uses Gaussian Laplace operator to extract the infrared light features and then fuses with the infrared lowfrequency information,the high-frequency part uses a fusion rule based on the detail enhancement of point sharpness to compensate for the shortage of low-frequency fusion.The fusion rule of detail enhancement based on point sharpness is used in the high frequency part to compensate for the shortage of low frequency fusion,and an adaptive factor is added to the low frequency fusion in order to adapt to different images.3.For the current visible and infrared image fusion algorithm is easy to ignore the thermal radiation information,a non-subsampling shearlet(NSST)based thermal radiation information retention algorithm is proposed to decompose the visible and infrared images in multiple scales and directions,and the fusion rules adopt the fusion rules based on thermal radiation information retention,and the fused high and low frequency subbands are inverted by NSST to generate the fused image,which makes the fused image can both The fused high and low frequency subbands are inverted by NSST to generate the fused image,so that the fused image can retain more thermal radiation information of infrared image and at the same time have good visual effect of visible light.Through comparison experiments with other eight algorithms and a variety of objective evaluation indexes,the proposed algorithm improves 0.1347,0.7441,0.7667,1.5898,0.9511,and 0.0747 on average than the other eight algorithms in six indexes of SSIM,PSNR,AG,SF,MI,and RMSE,respectively,which reflects the superiority of the proposed algorithm to more preservation of infrared thermal radiation information.