Figure 3: Classifier training for correlation-based approach

This figure depicts the development of a camera reference pattern in three steps. The reference pattern of a camera is determined by averaging the noise patterns from multiple images captured by the camera, and this is depicted in steps 1 and 2. The first step displays a set of five images taken with the same camera. Step two depicts noise extraction and averaging and displays a magnified portion of the extracted noise from each of the five images from the first step. The end result, step 3, is the camera reference pattern obtained by averaging the noise patterns from step 2, which serves as an intrinsic signature of the camera. It is represented as a single, shaded rectangle.