2009 IEEE Symposium on Industrial Electronics and Applications (ISIEA 2009), October 4-6, 2009, Kuala Lumpur, Malaysia Multi-sensor Remote Sensing Image Fusion Based On Retina-Inspired Model Hassan Ghassemian School of Electrical and Computer Engineering Tarbiat Modares University Tehran, Iran e-mail: ghassemi@modares.ac.ir Abstract—Remote sensing systems observe pixels in different portions of electromagnetic spectrum. These systems are designed within many competing constraints, among the most important being the trade off between the spatial resolution and the spectral resolution. To collect more photons and maintain image SNR, the multispectral sensors have a larger pixel compared to panchromatic sensors. With appropriate algorithms it is possible to combine these data and produce imagery with the best characteristics of both, namely high spatial and high spectral resolution. This process is known as a kind of data fusion. Some widely performed in the remote sensing community are HSI (hue-saturation and intensity) technique, PCA (principal component analyses) technique, and the Brovey transform technique. Recently, the Wavelet transform has been used for merging multi-resolution images. Normally, the objective of these procedures is to create a composite image of enhanced interpretability, but, those methods can distort the spectral characteristics of the multispectral images. This paper presents a multi-resolution data fusion scheme, based on visual channels image decomposition. This paper introduces a general issue of Retina-Inspired image analysis model, and application of the model in multispectral image fusion. A qualitative and quantitative comparison used to evaluate the spectral and spatial features performance of the proposed method with the others. Visual and statistical analyses show that the proposed algorithm significantly improves the fusion quality; compared to fusion methods including, IHS, PCA, Brovey, and discrete Wavelet transform (DWT). In this method, there is no need to resample images, which is an advantage over the other methods, it can perform in any aspect ratio between the panchromatic and MSS pixels. Keywords — Data Fusion; Multi-sensor; Remote Sensing; Retina-Inspired Model I. INTRODUCTION Earth observation satellites provide data in different portions of the electromagnetic spectrum at different spatial and spectral resolutions. For the full exploitation of increasingly sophisticated multi-source data, advanced data fusion techniques are being developed [1-3]. The fused images may provide increased interpretation capabilities and more reliable results. Many fusion methods have been proposed for fusing high spectral and spatial resolution data in order to produce multi-spectral images having the highest spatial resolution available within the data set. The higher resolution image is used as a reference which the lower resolution image is geometrically registered. Therefore the lower resolution image is up sampled to match the ground sample interval of the higher resolution image. The proposed method in this paper is a feature level image fusion technique. The objective of the multi-resolution image fusion is to generate hybrid high spatial resolution multi-spectral images that attempt to preserve the radiometric characteristics of the original low spatial resolution multi- spectral data. In this study, low-resolution multi-spectral MSS LandSat image fuse with a high-resolution panchromatic SPOT image to achieve optimal resolution in the spatial and spectral domains. Several methods have been used for this job, such as the IHS, PCA and DWT etc. [4-10]. The above methods and several other techniques have been developed to merge high-resolution panchromatic data with low-resolution Multispectral data. Normally, the objective of these procedures is to create a composite image of enhanced interpretability, but, those methods can distort the spectral characteristics of the multispectral images, and the analysis becomes difficult. The biological retina is more than a simple video sensor. It not only converts optical information into electrical signals but performs considerable processing on the visual signal before transmitting it to a higher visual system levels. Image fusion can incorporate the processing principles of human vision system [11]. This paper presents a multi-resolution data fusion scheme, based on retinal visual channels decomposition, motivated by analytical results obtained from "retina based image analysis, or multiscale image decomposition incorporates the visual channels phenomena" [12]. The energy packing the spectral features are distributed in the lower frequency subbands, and the spatial features, edges, are distributed in the higher frequency sub-bands. By adding the high-scale spatial features (extracted from a panchromatic image) to the low-scale spatial features (from TM images), the visual-channels procedure enhances the multispectral images. The retina model is based on Difference-Of-Gaussian (DOG) operator, which describes some of the receptive field properties of the ganglion cells [12]. 978-1-4244-4683-4/09/$25.00 ©2009 IEEE 500