|
The paper presents the results of merging lower-resolution spectral data (Landsat) with panchromatic images of higher resolution (IRS). The testing field (20 × 10km) covers the Cracow agglomeration. Thirty methods of merging satellite data (including their variants) have been tested in the research. They are based on canal substitution, on algebraic operations (including quotient operations), textural image features, look-up table, linear transformations, filters, local image transformations and on the analyses of various-resolution images. There are also combinations of the above approaches. Visual quality of the obtained synthetic images was assessed. The detailed analyses of different aspects of photo interpretation helped to maximise the objectiveness of the assessment. Object edges visibility and spectral distortion (colour change in a whole scene and/or in one object of specific spectral characteristic) were the analysed factors selected. They were farther divided into sub-factors and provided the basis to appoint the 30 merging methods categories. There emerged nine groups of various photo interpretation features. Four groups were described as useful, including one leading group of the most optimal photo interpretation features. Six colour composites were tested. Some methods with best parameters were indicated (HPF, IHS–HPF, generalised Laplacian pyramide, local correlation matching). Special attention was given to the methods which came low in the rank but have specific visual features (e.g. LUT) or have high spectral distortion which does not influence photo interpretation (e.g. IHS).
|