In spite of their high potential for automated discrimination between vegetation and human made objects, colour-infrared (CIR) aerial photos have not been in widespread use for traditional photogrammetric mapping. This is probably due to their awkward colour representation invalidating the visual analytical experience of the stereo analysts doing the actual registration of the topographical data. In this paper, we present a method for generating pseudo natural colour (PNC) representations from CIR photos. This enables the combination of automated vegetation discrimination with traditional manual mapping methods. The method presented is a dramatic simplification of a recently published method, going from a 7 step to a 2 step procedure. The first step is a classification of the input image into 4 domains, based on simple thresholding of a vegetation index and a saturation measure for each pixel. In the second step the blue colour component is estimated using tailored models for each domain. Green and red colour components are taken directly fron the CIR photo. The visual impression of the results from the 2 step method is only slightly inferior to the original 7 step method. The implementation, however, is much simpler, and the logical foundation more straightforward.
|Publication status||Published - 2005|
|Event||6th Geomatic Week Conference - Barcelona, Spain|
Duration: 8 Feb 2005 → 11 Feb 2005
Conference number: 6
|Conference||6th Geomatic Week Conference|
|Period||08/02/2005 → 11/02/2005|