5.  EXPERIMENTAL RESULTS

           The proposed coder has been run to encode several video sequences using 8x8 block size in ZTE coding. We used Daubechies’ 9-7 tap filter bank for all three decomposition levels in discrete wavelet transform. Gray format of original video frames has been obtained by luminance(called Y in YUV format) pixels. The experiments have been made using luminance components of the MPEG-4 test sequences “Akiyo”, “Bus”, “Silent” at CIF format(352x288 resolution).

           First result is that the proposed coder has managed to compress video sequences up to 90% (10:1). It is quite acceptable for a primitive video coder. Table 1 shows the results in different quality factors. The peak signal-to-noise ratio (PSNR) at 50% quality is 33.46 dB which has an acceptable view as can be seen in Figure 18.(b).

Table 2. Video sequence “Akiyo” first frame coding results.
 

(a)                                                                  (b)

Figure 18. Akiyo first frame, (a) original, (b) compressed with 50% quality.

 

           Second result is the cause of rapid PSNR fall in some video sequences. Videos that have big color distributions, having many bright colors, and big luminance values result rapid PSNR fall as quality factor goes down. This result can be seen in Table 3. Despite the quality factor having 70% in Figure 19.(b), the PSNR felt under 30 dB in sequence Bus.

Table 3. Video sequence Bus first frame coding results.
 

(a)                                                                  (b)

Figure 19. Bus first frame, (a) original, (b) compressed with 70% quality.
 

           However in sequences having small color distributions PSNR results fall smoother. This can be seen in Table 4. and the visual example in Figure 20.

Table 4. Video sequence Silent first frame coding results.

(a)                                                                  (b)

Figure 20. Silent first frame, (a) original, (b) compressed with 70% quality.

           Another result is the PSNR distribution in different frames at same quality. Akiyo sequence has been used to obtain PSNR values. Figure 21. shows the PSNR results for the first 50 frames in Akiyo. It has been found that in same quality, PSNR values does not change much.

Figure 21. PSNR results for Akiyo at 70% quality.

           Final result is about the compression ratio of video sequences. We reached the ratio of 10:1, which is not enough to compete with today’s compression schemes. In recent years MPEG-4 like compression schemes reach 60:1 of compression performance. By these schemes 10 MB of a video can be compressed down to 166 KB.