Image Regularization for Biometric Face Identification

Download PDF (694.41 KB)

Our biometric face ID system increases capabilities in the identification of faces from video sources. Current biometric face identification methods fail when used with low resolution images, for example, when faces are imaged either at a distance, or with a low-resolution imaging device, such as a cellphone or security camera. The complexity of images of moving human figures and faces presents additional problems for biometrics applications. This 'complex' motion cannot be adequately corrected with the current methods that use whole-image 2D affine transforms. The super-resolution approach overcomes this problem by aligning images locally, pixel by pixel. As a result, crisp edges, corners, and other features are much more clearly defined in the super-resolved version, despite complex motion in the video sequence.

To discuss licensing or collaboration activities, please contact MITRE's TTO.