Pansharpening of Clustered MS and Pan Images Considering Mixed Pixels

The component substitution (CS) scheme is one of the most efficient models used by different image fusion algorithms when merging multispectral and panchromatic images, acquired with different spatial and spectral resolutions. In this letter, a new CS-based image fusion method is proposed to reduce color distortion. First, multispectral images are clustered into several classes using spectral and spatial features, and then linear regression with non-negative coefficients is used to calculate summation weights for each class of pixels. To consider mixed pixels not belonging to any distinct class, the proposed method employs the fuzzy c-means algorithm. Qualitative and quantitative results are reported for two data sets, namely, Landsat-7 Enhanded Thematic Mapper Plus and QuickBird. Visual and statistical assessments show the validity of the proposed method.