blog




  • Essay / Image fusion technique based on PCA and fuzzy logic

    This article presents an image fusion technique based on PCA and fuzzy logic. The framework of the proposed image fusion technique is divided into the following main phases: Preprocessing phase Feature extraction based on principal component analysis Image fusion based on a fuzzy set Reconstruction of the final image Figure (1) shows the proposed image fusion framework and its phases.Fig. 1. The proposed approach to image fusion phasesA. Preprocessing PhaseThis phase includes three steps of registration, resampling and histogram matching. In the following1) Registration: Image fusion is the approach of combining two or more images of the same scene to obtain the most informative image. Image data is recorded by sensors on satellites; it may contain geometry errors which may be caused by the rotation of the earth during image collection. The images must therefore be saved. Registration is the preprocessing step in the fusion framework. Recording consists of superimposing two or more images of the same scene taken at different times or by different sensors. Registration is a crucial step in many image analysis tasks such as image fusion, change detection, etc. In this document; Ground control point technique is used to register the MS images to the Pan image as the reference image. The ground control point method is described as points on the earth of known location used as a georeference for the scene image. All MS images in this document are saved and the Pan image is used as the reference image; see Figure 2. (a) The MS image before recording. (b) The MS image after recording. Fig. 2. The impact of registration on satellite images2) Resampling: Resampling is a crucial step in the preprocessing of...... middle of paper ......CA is used to calculate the first analysis of components in order to redundant information and focus on pc1 which has the common spatial information in multispectral images. While the spectral information specific to each multispectral image is found on the other PCs. multispectral images are used as input data for PCA to obtain the pc1 which is used as input for the fuzzy set. Algorithm (1) shows the main steps of principal component analysis. Algorithm 1: The principal component analysis algorithm 1: Input: MS images (3 bands) in matrix form.%Perform PCA using covariance.2: data - MxN matrix of input data3: reshape 3 bands in 1*(m*n)4: subtract the mean5: calculate the covariance matrix6: obtain the eigenvalues ​​and eigenvectors of matrix covariance7: Recover the first principal component (PC1)8: Output: principal component (PC1, PC2 and PC3)