GAZE: An Attentive Processing Strategy to Detect and Analyze the Prominent Facial Regions

TL;DRAbstract

A new approach to a dynamic analysis of facial images is presented. Motivated by the eye movement strategies of the human visual system a computer based attentional mechanism is developed to recognize the prominent facial regions in natural human face images. The attentional mechanism is based on a feature representation, which is derived from graylevel face images by applying appropriate filter techniques. In a first processing step the most salient facial regions like the eyes, the nose, the mouth and the ears are sequentially localized. Subsequently, the detected and now spatially bounded regions are analyzed with more expensive methods. The exact positions of some relevant keypoints of these regions are determinable and a local interpretation of the detected image areas is derivable. By evaluating the different locally derived classification results a global interpretation of the whole face image can be generated. The presented attentive processing strategy is noted by a high degre

Chat with Paper

AI Agents for this Paper