Road is one of the most important and obvious extractable feature in satellite imagery. Automatic road extraction from satellite imagery has many advantages such as updating data bases by spending less time and cost. The aim of present research is the automatic extraction of forest roads map using Liss_IV sensor imagery of IRS_P6 satellite. Because of frequent irregular objects in forest, roads are very complicated for extracting automatically. Therefore, the designed methodology for this research was in a way that can deal with this problem. For this aim, image of the study area was classified into two road and non road areas by a fuzzy logic. Then, morphological mathematic algorithm was used to extract the existed roads. By this method, forest roads map was extracted automatically with 88% overall accuracy. Also, morphological mathematic algorithm showed a great ability for recovering road line that was hidden or was cut off under forest canopy.