Video sequence V2: change of expressions

The second video sequence we used is of a person changing expressions over time. The four expressions the face can switch to were:

• the neutral expression
• a smile
• a surprised expression
• a sad expression

The expressions changed several times through the video sequence. The sequence of expressions used was random, although some happenned more than others. This video is therefore less structured than the video V1.

The video has been shot with the same digital camera as video V1. It contains 3943 frames. Figure 4.9 shows some frames extracted from the sequence (see also examples/V2/V2_orig.m1v on the accompanying CD-ROM).

Figure 4.9: Frames extracted from the video V2.

Only 6 frames of this video have been marked up by hand using 43 points as depicted on figure 4.10. Extreme expressions have been marked up in the sequence and the appearance model was powerful enough to track effectively. The first mode of variation of the appearance model build from video V2 is shown on figure 4.11. The face has been successfully tracked through the video sequence. The relatives coordinates of the points as well as the pose, the scale and the position of the face have been reduced to 7 parameters for each frame.

Figure 4.10: Hand labelling of a frame from video V2.

Figure 4.11: First mode of variation of the appearance model extracted from video V2.

Figure 4.12 shows the synthesised version of the appearance parameters extracted from some frames of the video sequence V2.

Figure 4.12: Frames extracted from the video V2 after tracking.

The file examples/V2/V2_track.m1v on the accompanying CD-ROM shows the sequence of synthesised version of the tracked face for all frames from video V2.