Radiometric Adjustment of Aerial Imagery |
| | Abstract | The goal of this thesis is to develop a colour correction to a number of overlapping aerial photographs.
Maps created from aerial photographs are used for many practical purposes, for instance environmental investigations and city planning. In order to make maps of large areas, a mosaic from several photographs is created, and therefore the colours in the images should match to avoid visible seamlines between them.
COWI A/S has provided 22 overlapping orthophotos from aerial photos, which are used to investigate different methods of radiometric colour correction. Three different methods are investigated: Global histogram matching, global pixelwise matching, and global gradual matching.
In histogram matching the histograms in two neighbouring orthophotos are matched, and a linear transformation is estimated for each of the 22 orthophotos simultaneously in global histogram matching. Then global pixelwise matching, where linear transformations are estimated by simple pixelwise correspondence, is investigated. The third method described is global gradual matching, where the colour correction is performed under the assumption that there is a gradual change in colours over each orthophoto. In global gradual matching the results are improved by using boundary conditions and change detection.
Change detection is used to remove pixels that contain e.g. moving objects, or tall objects photographed from different angles, from the model estimation. In all three models a regularization term is added, such that a colour transformation, which is too large, does not occur.
In order to evaluate the quality of the results three measures are defined: The seamline measure, the saturation, and the contrast.
Experiments are performed to determine the optimal regularization, which show that it should be chosen as a trade-o, between making the seamlines less distinct, and obtaining a too low contrast for global histogram matching and global pixelwise matching. For global gradual matching the connection between the regularization of the model coe cients, the regularization used on the colour change in the boundary, and change detection is investigated.
The experiments show that the best results are obtained, when global gradual matching is used with boundary conditions and change detection. | | Type | Master's thesis [Academic thesis] | | Year | 2014 | | Publisher | Technical University of Denmark, Department of Applied Mathematics and Computer Science | | Address | Matematiktorvet, Building 303B, DK-2800 Kgs. Lyngby, Denmark, compute@compute.dtu.dk | | Series | DTU Compute M.Sc.-2014 | | Note | | | Electronic version(s) | [pdf] | | Publication link | http://www.imm.dtu.dk/English.aspx | | BibTeX data | [bibtex] | | IMM Group(s) | Image Analysis & Computer Graphics |
|