Enhancement And Fusion Algorithm Of Multi-resolution Analysis And Variational Satellite Imagery Based

In remote sensing applications such as visual interpretation, it is necessary to improve the visual quality of remote sensing images. Image enhancement can improve the contrast, enhance edge and restrain noise. It can make the enhanced image more acceptable in line with the visual characteristics. In order to provide the multispectral (MS) image with high spectral and spatial resolution simultaneously, pan-sharpening approaches inject the spatial details of the Panchromatic (Pan) band into MS bands. It can improve accuracy, reliability and comprehensiveness of understanding the target or scene context remarkably. It is also convenient for classification and identification.The main research work and innovations of this thesis include the following aspects:(1) The enhancement methods based on histogram equalization, Retinex and variational were analyzed elaborately. Histogram equalization is efficient to implement, but artifacts may occur in the enhanced images. The method based on Retinex is dependend on the setting of parameters. The ideal result by variational approach can be achieved, however, it has a high time complexity. Recently, the dark channel prior was applied to haze-removal and obtained prospective results. The feasibility of applying dark channel prior to remote sensing image enhancement was studied analytically.(2) An approach based on dark channel prior and bilateral filtering is proposed for the enhancement of remote sensing images. To address the high computational complexity of softmatting in the dark channel prior model, bilateral filtering is used to estimate the atmospheric veil, and the refined transformation map of He’s model is obtained. Crossing-color is observed when dark channel prior is applied to the enhancement of remote sensing images. Thus an improved algorithm for calculating transmission map is presented by improving the pixel values of depth image and making all pixel values not more than1. Compared to several image enhancement models based on Retinex and variational technique, the approach based on dark channel prior and bilateral filtering can increase image contrast effectively, and the results demonstrate the effectiveness of the proposed algorithm.(3) The fusion methods based on component substitution, multiresolution analysis and variational were reviewed and analyzed. The model based on component substitution has a high computation performance, but it may lead to spectrum distortion. The algorithm based on multiresolution analysis can achieve the high and low frequencies of the fused image at different resolutions. The varational method gives the energy functional according to the fusion requirements and it can attain a good trade-off between improving the spatial quality and maintaining the spetral signature. Because the modulation transfer function (MTF) of each band has a great effect on fusion result, the relationship between MTF and remote sensing image fusion result is analyzed.(4) To improve the fusion quality better, it is necessary to consider modulation transfer function (MTF) of the MS and Pan bands. A novel pan-sharpening model based on MTF and variational method is proposed. The energy functional consists of two terms. The first one is the spatial detail injection term, which injects detail information extracted from the Pan band with a high-pass filter into MS image. The second one is the spectral signature preserving term. The low-pass filter of’a trous’ wavelet is designed to maintain the multispectral information based on MTF of the MS band. The experimental results on QuickBird/IKONOS/GeoEye datasets demonstrate that pan-sharpening model based on MTF and variational method can obtain the fused MS image with high spectral and spatial quality. The fusion performance of the proposed model outperforms that of AWLP, IHS_BT, HPM-CC-PSF, NAWL and fast variational fusion model.