Facial Feature Extraction in an Infrared Image by Proxy With a Visible Face Image

Abstract

Thermal face image analysis has many applications, such as sensation evaluation and face recognition. Detection of a facial feature in an infrared (IR) image is an essential step in these applications. Unfortunately, it is difficult to directly extract some face features, even for salient features such as the corners of the eyes and mouth, and the nostrils, in the IR image because of the poor contrast between the features and the face in the IR image. On the other hand, face features can more easily be detected in visible images. This suggests that an effective algorithm is used to fuse visible and IR face images, which allows one to find facial features in IR images by finding them first in the visible counterpart. Current methods require a special apparatus to work in conjunction with the IR images. These methods are expensive and require elaborate calibrations. In this paper, a new method for detecting the facial features in an IR image by proxy, namely, to superimpose the visible face onto the target IR face, is proposed. We developed a method for the creation of photorealistic 3-D facial models from the stereo of a human subject. In our vision system, which includes visible and IR cameras, a 3-D geometric relationship between the IR and visible cameras is calibrated. Eye and mouth corners are detected in the visible image, and the head pose with respect to the visible camera is estimated based on the features. Then, the corresponding feature points in the IR image can be found by the head pose and the known 3-D geometric relationship between the visible and IR cameras. By doing so, skin temperature range within the IR image can be superimposed over the visible face image and vice versa. Experimental results on real images have verified the efficiency of the algorithm.