Main content area

A new pixelwise fusion based on feedback error correction

Naeimi Marandi, Reza, Ghassemian, Hassan
Remote sensing letters 2019 v.10 no.8 pp. 806-815
algorithms, data collection, remote sensing
In this letter, a novel pixelwise fusion is presented based on error correction. This method requires an initial step (initial fused image) for error correction. Thus, a new fusion framework based on discrete cosine transform (DCT) is proposed for error correction. This error correction includes the spatial and spectral parts. The spectral error is an error between the low-resolution multispectral (MS) image and the decimated fused image in each band. The spatial error is based on the Gram-Schmidt algorithm maximizing direction between the fused image and orthogonal vector of the panchromatic (PAN) image. One of the initial step (initials fused image) is the upsampling of the low-resolution MS image. Thus, this method is a pansharpening method independently. For the evaluation of the proposed method, two data sets are used in which the proposed method has better results both in quantity and quality aspects. The proposed method has the independence property from the dataset in as much as the error correction is done adaptively.