Infrared And Visible Image Fusion Research Based On Saliency Analysis

Image fusion usually refers to the combination of images collected by different sensors in a scene,the use of computers to process and fuse different images,the effective information in multiple images is retained in the final image.Infrared and visible image fusion is an important application direction in the field of image fusion.Visible light images rely on visible light reflected from objects for imaging,which is characterized by high resolution and contains a lot of detailed information.However,there are problems of poor imaging effect in harsh environments,such as smoke and fog.Infrared images rely on thermal radiation imaging on objects,and can also capture the target information in the scene in harsh environment,and the contrast is strong,but the separation rate is low,the background details are less,infrared and visible image fusion can make good use of the two image imaging characteristics,to obtain complementary information.When human eyes observe an image,they tend to first notice the regions with strong saliency,which often contain a large amount of effective information in the image.The purpose of image fusion is to extract the significant information in a variety of modal images.Therefore,image fusion method based on saliency analysis is the most consistent with the visual characteristics of human eyes.Latent low-rank representation can extract the corresponding saliency information from the image,and is often used in infrared and visible image fusion based on saliency.However,the traditional fusion results based on the latent low-rank representation have poor visual contrast in the target region,and the details are lost.To solve this problem,this thesis proposes a new fusion method based on the latent low-rank representation,which combines the latent low-rank representation and the ratio of low-pass pyramid to decompose the image,and obtains fusion rules with different design for the characteristics of each layer.Through the subjective and objective comparison with eight common fusion algorithms,it is proved that the fusion results of the salient target has a strong contrast,and a lot of details are preserved.Many existing fusion methods encounter the loss of salient object information,poor image quality of visible light source and low efficiency of algorithms.To solve these problems,this thesis proposes a fusion method of infrared and visible light based on salient region extraction and low-light enhancement.Different decomposition methods are designed for infrared image and visible image according to their characteristics,and appropriate fusion rules are designed.By comparing the fusion results and running time statistics on the images in the TNO dataset,the validity of the algorithm is proved.An improved low-light image enhancement algorithm proposed in this thesis is used to enhance the quality of the visible source image.Through comparison with other image enhancement algorithms and ablation experiments,it is proved that this enhancement algorithm can effectively enhance the quality of visible images.