Saliency Variation Based Quality Assessment for Packet-loss-impaired Videos
-
摘要: 针对网络中受丢包损伤的视频提出了一种基于视觉注意力变化的全参考客观质量评估方法.该方法基于视觉显著性检测在视频数据上的应用,考察受网络丢包失真影响的视频数据与标准参考数据在空间和时间上引起的视觉注意力变化,并根据此变化相应的视觉显著性在空间和时间上的差异,提出了一组客观质量评估方法.文中采用17个受丢包损伤的视频数据进行测试,并实施了主观评价实验作为评价标准.与传统的没有考虑人眼视觉显著特性的质量评估方法,以及目前主流的基于视觉显著区域/感兴趣区域对失真像素进行加权的方法进行对比,实验结果表明, 基于视觉注意力变化的方法较后两者与主观质量评估结果有更好的相关性, 能够更有效地评估丢包损伤视频的质量.Abstract: This paper presents a saliency variation based full-reference objective quality assessment metric for packet-loss-impaired videos. The method explores the application of visual saliency information. Motivated by the observation that packet-loss induced errors often change the spatial-temporal visual attention and correspondingly the saliency map, we explore the spatial changes in the saliency values between the original and distorted videos, and the temporal variation of the saliency map of the distorted video. The proposed metric was tested on 17 packet-loss-impaired videos and evaluated by a subjective test. Experiment results showed that the proposed method provides significant improvement in correlation with subjective results over traditional non-saliency quality metrics and the saliency/ROI (region of interest) weighted pixel-error quality measurements. This demonstrates that saliency variation is effective in evaluating the perceptual quality of videos affected by packet loss induces errors.
-
Key words:
- Saliency /
- attention change /
- video quality assessment /
- packet loss distortion
-
计量
- 文章访问数: 2237
- HTML全文浏览量: 93
- PDF下载量: 926
- 被引次数: 0