Computer graphics and computer vision are inverse problems. Conventional computer graphics begins with building geometric models of the scene and eventually produces an image, whereas traditional computer vision starts with input image (sequence) and generates geometric models. Lately, these two fields are seen to complement each other for creating and rendering models and scenes with greater realism and efficiency.

Though photorealistic renderings can now be achieved with traditional computer graphics, the process involved is rather tedious. Creating realistic models is a time consuming task and often requires modeling (artistic) skills and advanced techniques such as radiosity and global illumination, which tend to be very computationally intensive. Consequently, modeling tends to be hard, inefficient and time-consuming.

The process of recovering 3D structure from 2D photographs, now known as image-based modeling, has been one of the central endeavors in computer vision. Conversely, the process of rendering objects and scenes from images and geometry is known as image-based rendering which caters to the typical requirement of photorealistic rendering in computer graphics. Thus, this enables a way in which the modeling may not be a tedious task and the photorealistic rendering may not be slow or inefficient.