Research On Image Fusion Method Based On Target Segmentation And NSCT Transform

In recent years,multi-sensor image data fusion technology has been widely used in machine vision,remote sensing,medicine,military and agricultural fields.Image data fusion technology refers to the fusion of the same target scene images from different sensors or from the same sensor into a clear,more informative and conducive image with highlighting target.Along with the development of image fusion technology,many different algorithms have emerged,but they all have their own more applicable scene scope,which has certain limitations.Therefore,the fusion method proposed in this paper mainly focuses on infrared and visible images.The main work of this thesis includes:(1)A fusion method of infrared and visible images based on object segmentation is proposed.In this method,pedestrian information in infrared image is taken as target and visible image as background;infrared target image in infrared image is segmented by improved region growing algorithm;Finally,target image and background image are fused.Aiming at the selection of initial seed point and growth criterion,a method combining maximum density clustering algorithm(MDCA)with maximum brightness principle is proposed,which can realize the adaptive selection of initial seed point and growth criterion according to source image and improve the segmentation accuracy of infrared target.The experimental results show that the method not only retains the background information of the visible image,but also effectively adds the target information of the infrared image,and the image fusion effect has been improved.(2)An improved image fusion method based on non-downsampling contourlet transform(NSCT)is proposed.According to the characteristics of image high and low frequency components,the method uses principal component analysis and average gradient weighting method as the fusion strategy of low frequency subband.The adaptive selection method of regional variance and maximum value is used as the fusion strategy of high frequency subband.The study of fusion strategies provides a new way of thinking.Considering the advantages of NSCT transform withdirectionality and translation invariance,we use it for image decomposition and reconstruction.This method uses two different types of source images as experimental data.Compared with other fusion algorithms,it shows that the method is feasible and effective,and robust,the quality of the fused image is also improved.(3)Combined with the two methods mentioned above,a fusion method based on target segmentation and NSCT transform is proposed.The infrared region in the infrared image is segmented as the target image by using the improved region growing algorithm.The background image is obtained by the image fusion method based on the improved NSCT transform,and the target image and the background image are merged to obtain the fused image.The experimental results show that the method can obtain better fusion effect than the two methods mentioned above.