Fusion of very high-resolution UAV images with criteria-based image fusion algorithm

Abstract

Image fusion is the combination of spectral characteristics of a low-resolution image and spatial features of a higher resolution image to produce the spatially enhanced image. Different fusion techniques have been introduced in the literature. Fusion algorithms tend to yield good results when using single sensor images as input data sets. Fusion process becomes more challenging in case multisensor images are used. The spatial resolution ratio of the input images is another key factor which makes fusion process more challenging. A fusion algorithm is considered successful if it achieves to increase the spatial details without damaging the colour content of the low-resolution image. In this study, criteria-based fusion algorithm, introduced for the fusion of images with four bands, has been further modified for the fusion of a WorldView-2 MS (1.85 m) and an orthophoto panchromatic image (10 cm) produced with the images taken from an unmanned aerial vehicle (UAV). The spectral and spatial quality of the fused image produced with this algorithm has been evaluated qualitatively and quantitatively, and compared with those of advanced fusion methods such as Gram-Schmidt, FuzeGo, High-Pass Filtering, Ehlers, Hyperspherical Colour Space, Modified IHS and Adaptive Wavelet-Based algorithms. The results show that the criteria-based algorithm is very successful in keeping the colour content and gives satisfactory spatial detail enhancement compared to other algorithms.