Multispectral images fusion by a joint multidirectional and multiresolution representation

Bayesian super-resolution pansharpening using contourlets

“…We are working on the method that it can estimate the parameters automatically. The resulted images corresponding to the reconstruction of the synthetic image using the NSCT method in [4], the SR method in [5], and the proposed method are displayed in Fig. 1(d)-(f), respectively, and Table 1 shows the corresponding quantitative results.…”

“…1(e), introduced less details in the band 3 (blue) since the blue band contributes only a 10% to the PAN image and more into the band 2 since it has the highest contribution, a 60%, which is reflected as a greenish color near the edges of the image. The NSCT method in [4] incorporates details in all the bands but produces a noisy image, see Fig. 1(d).…”

“…We compared the proposed SR using contourlets method with the SR method in [5] and the additive NSCT method [4]. To assess the spatial improvement of the pansharpened images we use the correlation of the high frequency components (COR) [1] which takes values between zero and one (the higher the value the better the quality of the pansharpened image).…”

“…The panchromatic image contains the details of the high resolution MS image but lacks of its spectral information. Following the wavelet and contourlet based pansharpening methods, see [2,4], the PAN image is considered here as a combination of the high frequency details of the high resolution MS bands, plus a residual, low pass filtered image, which have an unknown relationship with the MS bands. The relationship between the PAN and the high resolution MS images is defined in this paper as…”

“…In the literature, this task has been addressed from different points of view (see [1] and [2] for a description and comparison of pansarpening methods). In recent years, Contourlet-based algorithms, specifically non-subsampled contourlet transform (NSCT) [3] based ones, are becoming popular [4]. The contourlet transform provides a complete shift-invariant and multiscale representation, with a fast implementation.…”