Research Of Multisensor Image Fusion Algorithms Based On Bandelet Transform

To suit some special requirements,multisensor image fusion methods fused images from different sensors through certain algorithms.Nowadays,multisensor image fusion has been widely used in medical,remote sensing,computer vision,target recognition,etc.Firstly,a systematic introduction about the image fusion principle,application and its development is presented.And the principle of bandelet transform is studied.Bandelet represents the geometrical regularity of image structure and sharp image transitions such as edges efficiently.In this thesis,these advantages of bandelet are used in image fusion to improve the fusion result.A multi-focus image fusion algorithm based on Bandelet transform is developed.Source images are firstly decomposed by Bandelet transform,then different fusion rules are used to the Bandelet coefficients.Finally the fused image is reconstructed by performing the inverse Bandelet transform.Experimental results indicate that the Bandelet-based fusion algorithm represents the edge and detailed information well,especially in images with abundant texture and edges information.Secondly,a novel image fusion algorithm based on nonsubsampled Contourlet transform(NSCT)and Bandelet transform is proposed in this thesis.As two different MGA tools,The NSCT can give an asymptotic optimal representation of edges and contours in image by virtue of its characteristics of good multi-resolution,shift invariance,and high directionality.The two transforms are jointed:first the image is decomposed into different frequency subbands by NSCT,and high frequency subbands are further decomposed by Bandelet transform.Next,high frequency subbands is fused and then reconstructed by using inverse Bandelet transform.Finally,both low and high frequency subbands are fused by inverse NSCT.Several different experiments are adopted to demonstrate that,results obtained based on our method contain more details information with smaller distortion than any other methods does.