Image fusion method based on total variation and àtrous wavelet

This work proposed a novel multiple sensors image fusion algorithm taking advantage of àtrous wavelet and total variation (TV). The àtrous wavelet is firstly used to decompose the multi-sensor source images into a low frequency subband and a series of high-frequency subbands. Then, fusion rules are applied to low and high frequency subbands respectively. The Block-principal component analysis (B-PCA) is introduced to estimate sensor gains in the low-frequency subband. Thus, a TV seminorm based approach is then used iteratively to obtain the fused image. The proposed approach can restrain the distortion which is introduced by B-PCA. Numerical simulations are carried out to validate our method.