We made 225 encoded sequences from 9 raw videos that have a length of 8 seconds, with a resolution of QCIF.
In order to evaluate the performance of the proposed NR-B metric, we firstly validate the proposed method with five QCIF test sequences as low quality (LQ) video, as shown in Fig.
3 shows the coded key frames in akiyo QCIF sequence by JVT-G012 and the proposed scheme at 32kbps.
6 show the PSNR and SSIM comparisons of the reconstructed akiyo QCIF sequence at 32kbps and reconstructed container CIF sequence at 128kbps by JVT-G012 and our scheme.
More experimental results between our proposed scheme and JVT-G012 for QCIF test sequences with different motion activity (low: container, grandma; moderate: foreman, news; high: carphone, salesman) at various target bit rate are shown in Table 1.
Rate-Distortion (R-D) of "Habour" sequence Dependency Temporal Quality Bit-rate PSNR QCIF
The "Foreman" CIF and "Highway" QCIF sequences are used as the video traces , and encoded, using H.
The video sequence was "Highway" in the QCIF format with 2000 frames .
11b without RTS/CTS RetryLimit 4 MAC bandwidth 11 Mbit IFQ length 50 Interface queue type Drop-tail priority Routing protocol NOAH Video trace "Foreman" CIF with 300frames Highway QCIF with 2000 frames Table 2.
Lower resolutions, such as QCIF
, QVGA, CIF, and VGA, are also supported.