Raster topographic images consist of a set of layers depicted in arbitrary color. There exist strong correspondence between the color of the layer and its semantic meaning. Often there is a need to separate or extract semantic layers from the map. The separation results in severe artifacts in places where semantic layers would overlap (e.g. elevations lines drawn on top of the topographic map). In the current work, we design the technique to reconstruct the sema ntic layers from the color layers resulting from the image separation process. The proposed technique pr ovides good visual quality of the reconstructed image layers, and can therefore be applied for se lective layer removal/extraction, which is often necessary in map processing and analyzing applications. It improves the accuracy of the data analysis and measurement tasks. The technique requires few co mputation resources and can be successfully used in mobile computers and terminals.