Facial asymmetry analysis based on 3-D dynamic scans

Abstract

Abstract-Facial asymmetry is an important characteristic used in a number of applications. It plays a vital role in human perception of attractiveness and as such has been used in psychology including research on facial expressions evaluation as well as in plastic surgery and orthodontics. It has been also recognized as a biometric feature used for identification and has important applications in detection and quantification of medical disorders, including neurological conditions, such as stroke or Bell's palsy. The current methods for assessing facial asymmetry mainly rely on manual semi-automatic procedures which have significant limitations as they are subjective in nature. This paper presents a computer-based methodology of automatic facial asymmetry analysis which has been used for detection and quantification of facial dysfunctions. The method is based on dynamic 3-D scans of human faces. Using facial sequences from the Hi4D-ADSIP database, the preliminary evaluation results suggest that the proposed method is able to assist in quantification and diagnosis of facial dysfunctions for neurological patients.

A Survey on Computer Vision for Assistive Medical Diagnosis From Faces

“…Methods based on analysis of the eyelid motion (blepharokymography) have also been developed to assess facial paralysis during blinking [106]. As an extended research on facial asymmetry using 3Ddynamics scans, Quan et al [107] showed that captured facial dysfunction can also discriminate stroke patients, also suggested by Matuszewski et al [108]. Similar study was performed using Kinect by Breedon et al 109.…”

Statistical shape modelling for expression-invariant face analysis and recognition

“…It is obvious that the areas of eyes, nose, and mouth corners have distinguishable values compared with other facial areas, even for different expressions. In combination with another facial landmark detection methods using the profile of facial symmetry plane [44], an additional six key landmarks can be identified that are located on the upper nose base, nose corners, upper and lower lip tips. Figure 3b demonstrates the location of all 12 key landmarks extracted from a 3-D face.…”

3-D Face Analysis and Identification Based on Statistical Shape Modelling

“…Using the approach proposed by Dorai and Jain [13], a set of six key landmarks can be extracted using the shape index: two inner eye corners, two outside eye corners and two mouth corners. In combination with another facial landmark detection methods using the profile of facial symmetry plane [14], an additional six key landmarks can be identified that are located on the upper nose base, nose corners, and upper and lower extremities of lip. Once key landmarks are located, more can be identified hierarchically using the geodesic-based sampling technique.…”

“…The Knearest neighbour method can also be applied to retain k edges for each x i and remove other connections. Derived from (14), the objective function can be rewritten as, Proposition. The objective function can be rewritten in the following short form where…”