Acquiring photos and videos has become a new norm in construction and building projects. However, imagery data is not utilized effectively due to the shortage of required skillsets in the industry and nonfamiliarity with classic image processing methods. Computer vision research in the context of construction and building has heavily focused on the interface between machine learning, and object tracking and activity recognition. Although positive results have been reported, namely improved productivity, safety and quality, implementations in the industry will not be immediate. Furthermore, algorithms such as convolutional neural networks (CNN), residual neural networks (ResNet) and recurrent neural networks (RNN) usually need to undergo extensive transfer learning in order to capture project-specific information in civil infrastructure engineering. This paper revisits classic image processing methods that can capture clues of site scenes with capability of high-level reasoning and inference. The work contributes to the body of knowledge by reviewing color, geometry and feature-based diagnostics in project environments.