Photo-simulations are illustrations which attempt to accurately simulate the visual appearance of something using photography, digital "paint" and 3D graphics.  The following article, originally written in 1993, describes the steps of the process. 

Before	Alternative A	Alternative B

The images are from a series of visual simulations produced for the Minnesota Department of Transportation to illustrate proposed highway design alternatives.   The simulations were used to explain the project design alternatives to the affected communities.

1) Photograph Existing Site Good site photography is one of the most important ingredients to a successful photosimulation. The base photograph used for the simulation needs to be chosen with special attention to the audience for the simulation and to the project issues being addressed. This aerial photo allowed us to cover the design layout, specific right of way impacts and general aesthetic issues all in one view.
2) Model Proposed Design Most constructed or manufactured features of the proposed design are modeled in Autodesk's 3D Studio directly from CAD data and paper drawings. The model is edited as a wire mesh representing 3D surfaces in the design. Most of the 3D mesh elements for this project were created by CAD technicians using Softdesk's civil design software for AutoCAD and imported to 3D Studio. Modeling is often more than half the effort involved in producing photosimulations.
3) Match Camera Perspective In 3D Studio, we can adjust the virtual "camera" until the perspective matches our base photograph. We start by placing a camera in the model at or near the location and altitude of the photographer when the base photograph was shot. We then add a few landmarks, visible in the photograph, to the 3D model. We adjust the camera's position and perspective until the modeled elements perfectly match the photograph.
4) Edit Materials Surface materials transform geometric 3D Studio meshes into photorealistic rendered objects. The 3D Studio materials editor allows us to modify the visible properties of our mesh objects. Materials can have scanned photographic textures, mapped patterns, reflection, transparency, multiple colors, shininess and bumpiness. To ensure the modeled elements of the photosimulation blend seamlessly into the background photograph, we sample colors and textures from the photograph as we develop our model materials.
5) Render 3D Model After we have placed lights in the model to simulate the lighting of the photograph, a high-resolution photorealistic rendering of the model is generated. This is the one part of the process where we just "hit a button" and wait for the computer to generate a picture for us. While test renderings generated through the course of developing the model may take only seconds for a computer to process, the final renderings used in photosimulations require between 15 minutes and 3 hours for processing on a 90 MHz Pentium with 64 Mb RAM.
6) Composite 3D Rendering with Photograph When the computer-generated rendering of the model is composited with the background photograph, the rendering looks close to being complete, but still requires a considerable amount of painting and retouching. Wherever there are foreground elements, such as buildings, in front of our 3D rendering, these will have to be painted back into the scene manually in the next step.
7) Paint and Retouch Final Photograph The final stage of creating a photosimulation is done using image editing software such as Adobe PhotoShop. Image editing tools allow us to cut and paste between photographs, adjust colors and "paint" onto the image. Organic details such as trees, landscaping and surface weathering are typically painted in using a digital airbrush. Vehicles and people are pasted into the scene by borrowing from other photographs. Painting is one of the most dynamic parts of the photosimulation process, but is also the most labor-intensive, typically representing 50% of the effort on a typical simulation.