Methods Of Multi-resolution Image Fusion With Under-water Sonar Image

Imagesonar is the detection device commonly used in marine activities through the imaging information.However,due to complexity of the seabed as well as influence of various environments and sonar itself,information is often a little by one collection.It is often necessary to collect the information of the same scene more than onec.Images from the sonars contain much complementary information.The fusion of the complementary information will get the image rich.As,the sonar iamge has the charateristic of low resolution,large noise,narraw gray distribution range and weak feature,sonar image fusion is rather difficult.This paper,based on the analysis of the existing theories and methods of image fusion,processes the corresponding methods from image registration to image multiscal decomposition and fusion strategy,so as to effectively solve the sonar iamge fusion problem.The main work of this paper is as follows:Firstly,the preprocessing operation before image fusion is studied including the methods of denoising and registration.Based on the analysis of the frequently used denoising methods,we choose suitable methods according to the number of noise.Then several methods of registration are introduced.According to the characteristics of the sonar image,a more suitable registration algorithm for sonar image fusion and higher registration precision is proposed.The corresponding experimental results show the effectiveness of the improved registration algorithm.Then,the framework based on multi-resolution decomposition is showed.and the characterisitic of multi-resolution decomposition in image processing is introduced briefly.This paper study the non-subsampled shear-wave transform detailedly.The fusion experiment applying the methods is carried out to verify that the fusion result is good when fusing the image without noise and is bad when fusing the image with much noise.Next,this paper proposes a multi-resolution method to suppress the noise while preserving the edge information of the image,according to the traditional multi-resolution decomposing fusion method which is applied in the noisy fused sonar image with the blurring edge.The non-subsampled Laplace transform in non-subsampled shear-wave transform is replaced by non-linear diffusion filtering.Additionally,In terms of low contrast and too much noise of sonar images,the corresponding fusion strategy is introduced.One strategy is about coefficient adjustment after multi-scale decomposition,aimine to suppressing the noise and preserving the edge information of the image.The other is the compution of significant valuebased on the local variance and local entropy of information avoiding the one-sidedness.Finally,comparative experiments are conducted to verify the effectiveness of the proposed multi-scale decomposition method from the way of multiscal decomposition and the fusion rules.The experiment results show that the proposed shearlet transform and the proposed fusion rules achieve effective results when applied on the sonar image with much noise.