Input Sequences

The video sequences used in the simulation were three. They are all QCIF size, 4:2:0 colour sampling, which means $176 \times 144$ pels for the luminance component Y, and $88 \times 72$ pels for each of the chrominance signals, C_b and C_r. All images are sampled at 15 Hz, that is, the frame rate is 15 fps.

Therefore, the sequences differ only at the content. The first, named `Akiyo,' is a typical single person head and shoulders image with medium spatial detail and low foreground motion with stationary background. The second, named `Mother & Daughter' or `M&D,' differs only at the point that motion is not restricted in a small region, as it features two subjects. The third, which will be the toughest test for the experiments, is the `Foreman' sequence, featuring medium foreground and low background motion, as the camera is unstable and finally pans away from the subject's face.

  

Figure 6.1: Original frames, (a) Akiyo frame 111, (b) M&D frame 22.

Sample frames of the first two sequences are illustrated in Figure 6.1, and for the Foreman sequence in Figure 6.4a, page . Note that these sequences are widely used by researchers for testing video coding algorithms, particularly, they are all MPEG4 test sequences.

Hosted by www.Geocities.ws