Similar to Pattern Analysis in purpose, but different in scope, is the topic Scene Analysis. The goal is to extract from camera images not only patterns (such as, for instance, the shape of a human being), but also high-level information about the whole scene. This might include any information about motion, or even a full 3D model of the scene, reconstructed from the 2D projection that has been captured by the camera.

Fig.: A particle filter scans for movements in this infrared camera image

Scene analysis almost always works on sequences of images; by comparing images that have been captured at different times, it can be determined which areas of an image are static and which ones might contain moving objects.

The first step usually is to register subsequent images, which means compensating any changes in the image that have been caused by moving objects or the movement of the camera itself. Only after this step can two images be meaningfully compared with regard to their content.

After that, motion of any in-scene objects can be detected, for instance by using a particle filter, and tracked over time.

Finally, depending on factors like scene type, camera types, and lighting, some geometrical information can be extracted from the scene.