The Research Of Multi-source Image Fusion Based On NSCT

Image fusion technology can effectively utilize the images information obtained by different sensors,and takes full account of the complementary information and redundant information between the images attained by the sensor so that it can get more accurate judgment and more reliable estimation than a single information source.It not only enhances the sharpness of image information,improves detail accuracy,but also satisfies the needs of human vision and expands the application range of information.Currently,Image fusion processing has been widely used in remote sensing image processing,map drawing,medical scan imaging and so on.Therefore,it is of great significance to carry out in-depth research on the fusion technology of multi-source images.Firstly,this thesis analyzes and discusses the existing fusion methods in detail.Image fusion can be divided into three levels based on pixel level,based on feature level and based on decision level.The pixel level operation including component substitution method?IHS,PCA?,multi-resolution method?DWT,NSCT?and intelligent analysis method,is simple and can be realized easily.In this paper,IHS,PCA,DWT and NSCT are compared and analyzed.Among them,the best performance and widely studied is the NSCT method.NSCT has the advantages of multi-scale,multi-direction and translation invariance,which can reduce the loss of detailed information in the fusion process and increase the information reduction degree of the fused image.However,NSCT still has the problem of higher time complexity,inability to adapt to the fusion rules?including the number of layers to be decomposed?and poor resistance to noise.Then,this thesis adopts fast NSCT?FNSCT?combined with gamma correction method to improve image fusion efficiency and enhance image contrast,and improves a fusion method based on non-down sampled contour wave.Two sets of multi-spectral images and panchromatic images of SPOT and TM were collected,and the fusion responses of the proposed approach are also compared with other five state-of-the-art fusion approaches;depicting the superiority of the obtained fusion results.At the same time,the simulation results show that the performance of the algorithm in the Q_MI?Q_AG?Q_S?SAM?ERGAS algorithm is 3%?15.7%?12.1%?14.4%and27%higher than the traditional NSCT algorithm.At the same time,the time consumption is reduced from 30%.The experiment proves that the ultra-high definition,the ultra-high retention of spectral information and the ultra-low time consumption can be guaranteed in the algorithm.Finally,this thesis has conducted in-depth experiments and research on medical images with high noise resistance requirements.The importance of image denoising and algorithm adaptation is found in the fusion of CT image and MRI image.This paper proposes a method of image fusion based on traditional NSCT combined with SWT and NSCT.The algorithm is also improved by 6%,9%,36%and 11%on the performance indicators AG,Q^S,MI and Q^AB\F,respectively.The results of medical image fusion can show that the algorithm's anti-noise ratio is increased 10%.It is obviously that the proposed algorithm not only preserves the original information of CT&MRI images,but also removes the influence of noise.Moreover,the image clarity and structure have been improved to a large extent.In summary,the improved NSCT algorithm for multi-source images is superior in simulation experiments and can achieve a good fusion effect.