Al-Anezi, Thamer Z.
Validation of the automatic tracking for facial landmarks in 3D motion captured images.
MSc(R) thesis, University of Glasgow.
Full text available as:
The aim of this study was to validate the automatic tracking of facial landmarks in 3D image sequences captured using the Di4D system (Dimensional Imaging Ltd., Glasgow, UK).
MATERIALS AND METHODS:
32 subjects (16 males; 16 females) range 18-35 years were recruited. 23 facial landmarks were marked on the face of each subject with a 0.5 mm non-permanent ink. The subjects were asked to perform three facial animations from the rest position (maximal smile, lip purse and cheek puff). Each animation was captured by a 3D stereophotogrammetry video system (Di4D). A single operator digitized landmarks on captured 3D models and the manual digitised landmarks were compared with the automatic tracked landmarks. To investigate the accuracy of manual digitisation, the same operator re-digitized 2 subjects (1 male and 1 female).
RESULTS & CONCLUSION:
The discrepancies in x, y and z coordinates between the manual digitised landmarks and the automatic tracked facial landmarks were within 0.5 mm and the mean distance between the manual digitisation and the automatic tracking of corresponding landmarks using tracking software was within 0.7 mm which reflects the accuracy of the method( p value was very small). The majority of these distances were within 1 mm. The correlation coefficient between the manual and the automatic tracking of facial landmarks was 0.999 in all x, y, and z coordinates.
In conclusion, Automatic tracking of facial landmarks with satisfactory accuracy, would facilitate the analysis of the dynamic motion during facial animations.
Actions (login required)