Multi-sensor Image Fusion Based on Statistical Features and Wavelet Transform

In this paper, we have proposed a novel feature based fusion algorithm which integrates the fusion technique of low pass and high pass wavelet coefficients. First we have decomposed source images into discrete wavelet decomposition coefficients. Thereafter, approximation coefficients are partition into n number of blocks and features are extracted from each block in the coefficients. Two same correspondence blocks features are then combined to make a feature matrix M. Finally, the weight is computed by applying the principle component analysis (PCA) on the feature matrix M to fuse the same corresponding blocks. After fusion of n blocks, they are combined to get final fused low pass coefficients. In the next phase, all the details or high pass coefficients are fused by computing the relative information of the coefficients and activity measure of each pixel in the neighborhood. Finally, the new coefficient matrix is obtained by concatenating fused approximation and details coefficients and the fused image is reconstructed using inverse wavelet transform. To verify the superiority of our proposed algorithm, we have tested it on 100 pair different sensor images collected from Manchester University UK databases. Finally, the proposed algorithm is compared with one existing algorithm and it performs better in terms of subjective and objective perception.