Color image segmentation using false colors and its applications to geo-images treatment: alphanumeric character recognition

In this work an approach is proposed to segment the alphanumeric characters that present in a cartographic color map, using the model-of color RGB. Corresponding raster image is obtained by means of scanning, following to the strategies proposed in S. Levachkine et al. (2000). Our approach does not require a preprocessing of the images, because they only maintain the pixels that really belong to the characters we wish to segment, eliminating all those pixels that are not of interest, produced by a noise or obtained due to erroneous selection of scan parameters (for example, scan resolution), etc. The following identification of the alphanumeric characters supports by a set of neural network and the dictionaries with the character names related to the map or particular application that has previously been prepared.