Efficient image fusion method based on the Fourier transform by introducing sensor spectral response

摘要

Different fusion methods were proposed in image sensing applications to inject features extracted from the panchromatic image (PAN) with high spatial resolution into the multispectral image (MS) while trying to improve its low spatial resolution to reach that of the PAN image and preserve its spectral signatures. The objective is to obtain an image that should be observed by a sensor, which would deliver, if it existed, a single image with both the high spatial resolution of the PAN image and the high spectral resolution of the MS image. These methods, by injecting the same energy of the extracted features into MS image bands, result in spectral distortions. Since the PAN image is rich in frequencies, an interesting approach is to inject its high spatial frequencies into MS image bands. A new approach to image fusion using the Fourier transform with a new injection model defined from the sensor characteristics to avoid the injection of the same energy of high spatial frequencies into all MS image bands is proposed. A set of examples with both qualitative evaluation and various spectral and spatial quantitative measures are given to show the effectiveness and efficiency of the proposed approach. Also, this method is computationally fast and can be used to quickly merge a massive volume of data.