In emergency scenarios, immediate reconnaissance efforts are necessary. These efforts often take months to complete in full. While underway, building occupants are unable to return to their homes/businesses, and thus, the impact on the society of the disaster-stricken region is increased. In order to mitigate the impact, researchers have focused on creating a more efficient means of assessing the condition of buildings in the post-disaster state. In this paper, a machine vision-based methodology for real-time post-earthquake safety assessment is presented. A novel method of retrieving spalled properties on reinforced concrete (RC) columns in RC frame buildings using image data is presented. In this method, the spalled region is detected using a local entropy-based approach. Following this, the depth properties are retrieved using contextual information pertaining to the amount and type of reinforcement which is exposed. The method is validated using a dataset of damaged RC column images.