The paper presents the image processing method for detection forest regions from digitized old maps. The proposed scheme consists of two principal steps. In the first one the input color image is filtered using averaging filter and binarized by thresholding in the HLS color space. The second step is based on morphological image processing of binary images and results in smoothening of contours of forest regions and removing unnecessary objects. As experiments are showing, the proposed method produces correct results of segmentation, allowing extraction of forest regions that can be further effectively processed and analysed.