Research On Fusion Method Of Visible Light Image And Infrared Image

With the continuous development of sensor technology,there are more and more ways to obtain information.Therefore,information fusion technology is an inevitable development trend.Image fusion as an important branch has also received extensive attention.Due to the different imaging mechanism of infrared image and visible light image,the image contains information with certain complementarity,and the fusion of infrared and visible images can effectively improve image resolution and information.Based on the research of domestic and foreign research status,this paper proposes a fusion method of visible light and infrared image with detailed information and significant visual region.The method includes fusion method for image detail information and fusion method for image visual significant region.For the fusion of image detail information,this paper proposes an image fusion algorithm based on Convolutional Neural Networks(CNN)and Discrete Cosine Transform(DCT).In this paper,the image decomposition is first decomposed into the base layer and the detail layer by the guide filter.Secondly,the CNN and DCT are used to fuse the image detail layer,and then the DCT is used to fuse the image base layer.Finally,the merged base layer and detail layer are used to reconstruct Get a fused image.The advantage of this method is that a binary image containing the original image contour information is used as a guide image for guiding filtering,thereby decomposing into more fine detail information in the detail layer,and improving the contour blurring phenomenon in the infrared base layer.Then according to the characteristics of CNN,a fusion rule is redesigned to achieve a good fusion of image detail information.Aiming at the visual saliency region fusion of images,this paper proposes an image fusion method based on Visual Attention Model(V AM).The principle is to first obtain the image edge texture information parameters,so as to achieve the optimal selection of the VAM morphological features;Secondly,normalize the original image to generate a standard map;Finally,use VAM to obtain a visual saliency map,and visual Enhance the saliency map to achieve effective fusion of visible light images and infrared imagesThe experimental results show that this paper proposes a fusion method based on detailed information and visually significant regions to obtain fusion results compared with other typical fusion methods.It not only has better definition in subjective vision,but also has better performance in objective evaluation parameters.which performed.