AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2...

22
AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 Lin Liang 2 Jian Sun 2 Yangsheng Wang 1 1 Institute of Automation Chinese Academy of Sciences, Beijing, C 2 Microsoft Research Asia Beijing, China CVPR 2010

Transcript of AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2...

Page 1: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

AAM based Face Tracking with Temporal Matching and Face SegmentationMingcai Zhou1 、 Lin Liang2 、 Jian Sun2 、 Yangsheng Wang11Institute of Automation Chinese Academy of Sciences, Beijing, China2Microsoft Research Asia Beijing, China

CVPR 2010

Page 2: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

2

Outline• AAM Introduction• Related Work• Method and Theory• Experiment

Page 3: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

3

AAM Introduction• A statistical model of shape and grey-level appearance

Shape model Appearance model

Page 4: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

4

Shape Model Building:mean shape

:shape bases

,shape parameters

learn by PCAgenerate mean shape 、shape bases

Page 5: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

5

Texture Model Building:mean appearance

:appearance bases

:appearance parameters

W(x)

灰階值

Shape-free patchMean shape

Page 6: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

6

AAM Model Building

Page 7: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

7

AAM Model Search• Find the optimal shape parameters and appearance parameters to minimize the difference between the warped-back appearance and synthesized appearance

( , )W x p

p

( ( , ))I W x p

map every pixel x in the model coordinate to its corresponding image point( , )W x p

0s

Page 8: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

8

Problems- AAM tracker• Difficultly generalize to unseen images• Clutterd backgrounds

Page 9: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

9

How to do?• A temporal matching constraint in AAM fitting -Enforce an inter-frame local appearance constraint between frames• Introduce color-based face segmentation as a soft constraint

Page 10: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

10

Related Work -feature-based (mismatched local feature) Integrating multiple visual cues for robust real-time 3d face tracking, W.-K. Liao, D. Fidaleo, and G. G. Medioni. 2007 -intensity-based (fast illumination changes) Improved face model fitting on video sequences, X. Liu, F. Wheeler, and P. Tu. 2007

temporal matching constraint

Page 11: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

11

Method and Theory• Extend basic AAM to Multi-band AAM– The texture(appearance) is a concatenation of three texture band values• The intensity (b)• X-direction gradient strength (c)• Y-direction gradient strength (d)

Page 12: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

12

1. Select feature points with salient local appearances at previous frame2. I(t−1) to the Model coordinate and get the appearance A(t-1)

3. Use warping function W(x;pt) maps R(t-1) to a patch R(t) at frame t

Temporal Matching Constraint

Page 13: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

13

Shape parameter Initialization

,

Face Motion Direction

Page 14: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

14

Shape parameter Initialization

When r reaches the noise level expected in the correspondences, the algorithm stops

Page 15: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

15

Shape parameter Initialization-Comparison

Motion direction

Feature matching Previous frame’s shape

Page 16: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

16

Face Segmentation ConstraintWhere are the locations of the selected outline points in the model coordinate

{ }kx

Page 17: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

17

Face Segmentation Constraint-Face Segmentation

Page 18: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

18

Face Segmentation Constraint

Page 19: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

19

Experiments

Lost frame num

Page 20: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

20

Experiments

Page 21: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

21

Conclusion─ Our tracking algorithm accurately localizes the facial components, such as eyes, brows, noses and mouths, under illumination changes as well as large expression and pose variations.

─ Our tracking algorithm runs in real-time . On a Pentium-4 3.0G computer, the algorithm’s speed is about 50 fps for the video with 320 × 240 resolution

Page 22: AAM based Face Tracking with Temporal Matching and Face Segmentation Mingcai Zhou 1 、 Lin Liang 2 、 Jian Sun 2 、 Yangsheng Wang 1 1 Institute of Automation.

22

Future Work─ Our tracker cannot robustly track profile views with large angles

─ The tracker’s ability to handle large occlusion also needs to be improved