Academia.eduAcademia.edu

Figure 3 – uploaded by Jose Joskowicz

See full PDFdownloadDownload figure
Table 3. MOS to perceived quality relation

Table 3 MOS to perceived quality relation

Related Figures (14)

The basic video quality J. for each bit rate b (the video quality due to encoding degradation only, without packet loss) is expressed in Equation (2). where b is the bit rate and v3, vz and v; are three coefficients of the model.
The basic video quality J. for each bit rate b (the video quality due to encoding degradation only, without packet loss) is expressed in Equation (2). where b is the bit rate and v3, vz and v; are three coefficients of the model.
Combining Equation (1) and (2), the video quality V, without packet loss is expressed in Equation (3). According to the ITU-T Recommendation G.1070, coefficients v3, v4 and vs are dependent on codec type, video display format, key frame interval, and video display size, and must be calculated using subjective video quality tests. Provisional values are provided only for MPEG-4 in QVGA (Quarter VGA, 320 x 240 pixels) and QQVGA (Quarter QVGA, 160 x 120 pixels) video formats. In [9] a new set of parameters for the MPEG-2 codec are proposed.
Combining Equation (1) and (2), the video quality V, without packet loss is expressed in Equation (3). According to the ITU-T Recommendation G.1070, coefficients v3, v4 and vs are dependent on codec type, video display format, key frame interval, and video display size, and must be calculated using subjective video quality tests. Provisional values are provided only for MPEG-4 in QVGA (Quarter VGA, 320 x 240 pixels) and QQVGA (Quarter QVGA, 160 x 120 pixels) video formats. In [9] a new set of parameters for the MPEG-2 codec are proposed.
where V, represents the video quality determined by the codec distortion, b is the bit rate (in Mb/s), a is a constant that depends on the display format and vy and vs are other model parameters with a strong dependence on video content. Video clips are classified in three categories, according to the subjective movement content, and the model parameters v, and vs are calculated for each category (High, Medium and Low movement content).
where V, represents the video quality determined by the codec distortion, b is the bit rate (in Mb/s), a is a constant that depends on the display format and vy and vs are other model parameters with a strong dependence on video content. Video clips are classified in three categories, according to the subjective movement content, and the model parameters v, and vs are calculated for each category (High, Medium and Low movement content).
As can be seen in Figure 1, all the clips have better perceived quality for higher bit rates, as expected. In MPEG-2, in SD, for bit rates higher than 6 Mb/s all the clips have an almost “perfect” perceived quality (MOS higher than 4.5). At 3 Mb/s all the clips are in the range between “Good” and “Excellent”. However for less than 3 Mb/s the perceived quality strongly depends upon the clip content. For example at 2 Mb/s, MOS varies between 3.6 and 4.8, and at 0.9 Mb/s MOS varies between 1.9 (between “Bad” and
As can be seen in Figure 1, all the clips have better perceived quality for higher bit rates, as expected. In MPEG-2, in SD, for bit rates higher than 6 Mb/s all the clips have an almost “perfect” perceived quality (MOS higher than 4.5). At 3 Mb/s all the clips are in the range between “Good” and “Excellent”. However for less than 3 Mb/s the perceived quality strongly depends upon the clip content. For example at 2 Mb/s, MOS varies between 3.6 and 4.8, and at 0.9 Mb/s MOS varies between 1.9 (between “Bad” and
Table 4. Source video clips used Table 5. MPEG-2 and H.264 coding parameters used
Table 4. Source video clips used Table 5. MPEG-2 and H.264 coding parameters used
Figure 1. Perceived Quality as a function of the Bit Rate and video content
Figure 1. Perceived Quality as a function of the Bit Rate and video content
Table 6. vz and vy; values that best fits to the actual NTIA curves and Avg SAD/pixel
Table 6. vz and vy; values that best fits to the actual NTIA curves and Avg SAD/pixel
Table 7. Movement content vs SAD Figure 2 shows the relation between v4 and average SAD per
Table 7. Movement content vs SAD Figure 2 shows the relation between v4 and average SAD per
Figure 2. Relation between vy, with respect to SAD where s is the original video average SAD per pixel. The best values for c;.. cs were calculated (using least squares method) and are presented in Table 8.
Figure 2. Relation between vy, with respect to SAD where s is the original video average SAD per pixel. The best values for c;.. cs were calculated (using least squares method) and are presented in Table 8.
Figure 3. Relation between v; with respect to SAD
Figure 3. Relation between v; with respect to SAD
Table 8. MPEG-2 and H.264 coding parameters used Using Equation (6), the video quality estimation presented in
Table 8. MPEG-2 and H.264 coding parameters used Using Equation (6), the video quality estimation presented in
Table 9. Original and proposed model comparison
Table 9. Original and proposed model comparison
Figure 4. MOS dispersion in proposed model The MSE between the values derived using Equations (4) and (6) and the perceived MOS values is 0.015, meaning a very low “distance” between the model and the standard VQM.
Figure 4. MOS dispersion in proposed model The MSE between the values derived using Equations (4) and (6) and the perceived MOS values is 0.015, meaning a very low “distance” between the model and the standard VQM.
Figure 5. MOS dispersion in G.1070 original model
Figure 5. MOS dispersion in G.1070 original model
Related topics:
Computer ScienceDesignVideo QualityVideo CodecsVideo Signal Processing
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved