This paper presents a novel framework based on image processing techniques for the radio environment characterisation. Specifically, after digitalising a given scenario in accordance with the frequencies that are detected in each point, an image can be built whose pixel intensities capture the radio-electrical conditions. Through non-linear filtering and object detection operations the proposed framework allows identifying the homogeneous regions where certain frequencies can be detected. This methodology can be used in different contexts, such as the construction of databases to provide the terminals with spectrum awareness in flexible spectrum scenarios to facilitate the switch-on procedure or the secondary usage of spectrum, the support to devise proper strategies for cell and RAT selection, etc.