Log-Gabor directional region entropy adaptive guided filtering for multispectral pansharpening

The fusion of multispectral (MS) image and panchromatic (PAN) image, also known as MS pansharpening, has become an important research field in remote sensing image processing. However, the MS pansharpening method based on detailed information injection has two problems that have not been well solved: one is how to extract as many effective texture details from the PAN as possible, and the second is how to effectively inject these texture details into the MS, that is, how to avoid spectral distortion while maximizing the spatial resolution of the MS. To address these problems, in this paper, a novel log-Gabor directional region entropy adaptive guided filtering algorithm is proposed for MS pansharpening. The innovations are as follows. (1) A directional region entropy measurement index in the log-Gabor transform domain is defined to measure the strength of local texture structural information. (2) A directional region entropy guided image filtering (DEGIF) model is proposed, which overcomes the halo effect of the local region filtering model and improves the accuracy of extracting textural details to a certain extent. (3) A decision supervision template with information injection and an adaptive weighting scheme based on gradient entropy are constructed. The former ensures the accuracy of the information injection location, and the latter prevents distorting the spectral information after the textural details are injected. (4) A new MS pansharpening scheme is proposed, which can effectively improve the spatial resolution of the fused image while maintaining its spectral property. The results of comparative experiments show the superiorities of the proposed algorithm with regard to other state-of-the-art MS pansharpening approaches.