Image Fusion Based On Adaptive Pulse Coupled Neural Network And The Discrete Fractional Random Transform

With the development of modern remote sensing technology, more and more multi-resolution, multi-temporal and multi-spectral remote sensing image data are provided by different kinds of remote sensors. However, these remote sensing image data are not same in the different fields because of observation limit and dissimilar design of the remote sensors. As an important aspect of image fusion, the remote sensing image fusion investigates how to make use of image information from different remote sensors synthetically, which can get a deeper understanding of the same thing or target comprehensively, objectively and essentially.Recently, a new method has been proposed in the literature for the remote sensing image fusion by use of discrete fractional random transform. In the discrete fractional random transform domains, high amplitude spectrum and low amplitude spectrum components carry different information of original images. The high amplitudes carry the spectral information and the low amplitudes just carry the spatial details. If we fuse images in discrete fractional random transform domains, it has a lower influence on the same-strength changes in spectrum which guarantees low spectral distortion. What’s worse, the method can’t extract the high and low amplitudes with high efficiency and intelligence, which is inconvenient for practical applications.In order to meet the requirements of high spatial resolution and low spectral distortion, this paper presents a new approach for the remote sensing image fusion, which utilizes both adaptive pulse coupled neural network and discrete fractional random transform. In the proposed scheme, the multi-spectral and panchromatic images are converted into the discrete fractional random transform domains, respectively. We take full advantage of the synchronization pulse issuance characteristics of pulse coupled neural network to extract the high amplitude spectrum and low amplitude spectrum components properly, and give us the pulse coupled neural network ignition mapping images which can be used to determine the fusion parameters.This paper also proposes the discrete fractional random transform, which can make the spectrum distributed randomly and uniformly. Subsequently, we introduce this new spectrum transform into the image fusion field and present a new approach for the remote sensing image fusion, which utilizes both adaptive pulse coupled neural network and the discrete multi-parameter fractional random transform. In the image fusion of discrete multi-parameter fractional random transform domain, it can maintain most of the desired properties of discrete fractional random transform. In addition, the discrete multi-parameter fractional random transform is very sensitive to parameters with considerable robustness and security, which enhances the practicability in information security field. And the whole fusion process is very similar to the method which uses the both adaptive pulse coupled neural network and the discrete fractional random transform. A lot of numerical simulations are performed to demonstrate that those two proposed methods are more reliable and superior than several existing methods based on IHS, PCA, discrete multi-parameter fractional random transform etc. Not only have they better performance in preserving the spectral components of the remote sensing image, but also they can increase the spatial details of the fused image.