Research On Application Of Visual Saliency In Image Fusion

Image fusion is a processing method of merging several images obtained by different modes in the same scene or the same sensor at different times or modes into one image.Since a single image sensor is targeted at certain scenes,it does not reflect multiple features of the image.However,if the amount of data formed by images acquired by a plurality of image sensors is too large,it is necessary to extract and synthesize useful information in each image.Visual salience can reflect the ability of the target to attract visual attention in the scene.If the target image is analyzed and extracted significantly,the target area in the image can be enhanced to achieve the purpose of improving the fusion effect.Therefore,it is necessary to conduct in-depth research on the application of visual saliency in image fusion.Based on the relevant theories and models of visual saliency,this paper has carried out in-depth research on the application of visual saliency in infrared and visible image fusion and multi-focus image fusion.The main research contents are as follows:1.For the halo phenomenon of the Guided Filter Frequency Tuned(FTG)method,the method used in this paper is to calculate the saliency of the pixel in the local area to determine the position of the pixel,thus obtaining the weighted state of the regularization parameter.So that the improved FTG algorithm can weaken the original guided filtering halo phenomenon.2.Aiming at the problems of blurred details of infrared image,poor contrast and unclear visible light texture,the improved FTG algorithm is used to extract the infrared image saliency map,and then the original image is subjected to Non-Subsampled Contourlet Transform(NSCT).The saliency area of the low frequency part of the image is in accordance with the fusion rule of adaptive weight,and the non-significant area is according to the fusion rule of regional energy and regional definition;the high frequency part of the image adopts the absolute value of the coefficient to take the large method.This algorithm can better improve the problem of losing visible background information in the fusion of infrared and visible images and make the target area of the infrared image clearer.3.Aiming at the problem of unclear area in multi-focus image,the improved saliency algorithm is used to extract the saliency map of multi-focus image,and the sparse decomposition feature is used to reduce the complexity of NSCT algorithm and improve the fusion efficiency.The low frequency part of the image uses the significant size to guide the sparse coefficient,the high frequency part compares the Laplacian sum,and finally the multi-scale inverse transform is used to obtain the fused image.This algorithm can achieve better fusion effects than using multi-scale or sparse fusion algorithms alone.