Integration of Panchromatic and Multispectral Images by Local Fractal Dimension

Resumen

The fusion image strategies are a good solution to obtain a synthetic image with high spatial and spectral characteristics simultaneously. Some of them are based on the Wavelet Transform, computed by means of the à trous algorithm (AWT). Most of them do not differentiated between spectral bands. In this sense, a new approach that weights differently the spatial information integrated from the high resolution image in each of the fused image spectral bands by the optimization of the trade off between the spatial and spectral quality of the fused images, was proposed. The main problems of this approach are that a unique weighting factor for the whole spectral band is computed, and the need of indices, that separately measure the spectral and spatial quality of the fused images. In this work, a new strategy that tries to avoid the problems above mentioned is introduced. For that, it is proposed to determine a local weighting factor for each panchromatic pixel by means the fractal map, using the box-counting algorithm. Panchromatic and multispectral Quickbird images have been used to show the performances of this new methodology. The local quality of the final fused images has been evaluated by means of local quality maps of Q index. It has been proved that the proposed fusion strategy preserve the high frequency information of the panchromatic image in areas with a high detail, while in homogeneous areas the low frequency information of the multispectral image are conserved.