Image fusion is a process by which

Image fusion is a process by which a new image is created which combined two or more images. The main aim of such fused images is to retain as much source information as possible and it is expected that the fused image will be better than, or at least as good as, the performance of input images. Fig 3 shows the general block diagram of the fusion method. Image fusion techniques can be classified as: pixel based, feature based and decision based. The most commonly used method involves arithmetic computation technique such as Bovey Transform (BT) and Synthetic Variable Ratio (SVR).But the method suffers from the drawback that it is not efficient to quickly merge such large volumes of satellite data.Recently there are some geometric analysis tools which are applied for image fusion based on transforms such as curvelet transform, contourlets transform, wedgelet transform and Non Subsampled Contourlet Transform (NSCT). For the satellite images, it is important to analyze its geometric structure .For this, the Shift-Invariant Shearlet transform (SIST) is proposed for remote sensing image fusion [7]. As the first step the feature vectors of multispectral and panchromatic images are extracted and then they are divided into different regions using Fuzzy C Means (FCM). The SIST gives the representation of the first principal component of the multispectral image. The principal component in an image is obtained using entropy component analysis and panchromatic images. Various model based methods are also being used in the area of image fusion. One such method is a hierarchical Bayesian model to fuse multiple multi-band images. Image fusion based on image decomposition and sparse representation is proposed in [8]. Here the input image is divided into cartoon and texture components. The fusion of Method Feature Extracted Limitations 1. Multi image saliency analysis[4] ROI extraction such as clouds Presence of unwanted background information 2. Uniform Competency Feature Extraction[5] Rotation and scale invariant Local features Varying matching results for different data 3. Digital Surface Models [6] Pixel and feature level extraction of urban scenes Difficulty in multi feature classification 4. Reversible jump Markov chain Monte Carlo sampler[7] Extraction of rivers, channels and roads Sensitivity to experimental settings Author Segmentation Advantages 1.S. Suresh and S. Lal (2016)[6] Cuckoo Search,McCulloch’s method Computationally efficient and better convergence 2. I. Grinias , C. Panagiotakis , G. Tziritas (2016)[7] Markov Random Filed method Better classification accuracy with less visual and shape features 3. D. Marmanis et al.(2017)[8] Deep convolutional Neural Network Efficient boundary extraction improves classification 4. S. Pare , A.K. Bhandari , A. Kumar , and G.K. Singh(2017)[9] Levy flight firefly algorithm Good level of multilevel segmentation with minimum computation time 122 the cartoon components is carried out using a spatial based fusion.