The availability of maps varies. In Finland, the situation is very good and maps are available in a wide variety of formats from almost every part in Finland with a very detailed scale (1:20 000). Map formats can be generalized into the following main types (figure 4):
· Vector maps
· Color raster images
· Gray-scale raster images
· Binary maps
 

Figure 4: Different ways of representing a map: color (left) and gray scale image produced by combining different semantic components, and the same components plotted as binary images (right). 
Click on any of the pictures to get an enlarged view.
 

Firstly, we need to implement conversion from vector format to raster so that the information is divided into semantic layers (one color per layer for example). Secondly, we need to convert also color and gray-scale images into binary layers. Figure 5 illustrates that we can always make straightforward conversion from color to gray-scale, and then divide the image to bit-planes. This approach, however, is not very efficient and it would be better if we could segment the image into the layers directly according to the color information. If we have clean original image, this can be done rather easily. If the image, however, contains scanning noise, dithering or it has been compressed by lossy methods such as JPEG, then the problem is more difficult and requires intelligent pre-processing in order to maintain the quality and keep the storage size compact.

During the project, we first implement the straightforward solutions so that we can convert any image into the MISS format. Next we will implement support for ArcShape-format so that we can get compressed MISS maps directly without intermediate processing stages. Finally, we develop better methods for the direct conversion from color images. In principle, it is a question of color segmentation but contains also other sub problems: how to solve the number of color segments, deciding the colors for the segments, and removal of noise. The problem is that we may not have any information about the original semantic separation and we can only estimate it. This part of the project is demanding research problem and the result depends how much we manage to solve the other tasks first, and how much resources can be focused on this problem.

Figure 5: Necessary conversions between different image types.
Click on the picture to get an enlarged view.
 
 
 

<<<<goals

next>>>>