作　　者： ; ; ; ; (张代军）;

机构地区： 深圳信息职业技术学院

出　　处： 《深圳大学学报（理工版）》 2013年第5期462-468,共7页

摘　　要： 提出一种基于时空显著图的可伸缩视频码率控制方法.建立数学模型描述人眼视觉注意机制中显著区域提取,引入注意焦点转移等视觉感知任务,从时空域角度计算获取显著图.根据时空显著图提出基于贪婪算法的增强层码率控制方法实现码流提取,结合人眼中央凹视觉偏离角-感知清晰度关系特性进行码率分配优化.实验结果表明,该方法能有效模拟人眼视觉注意特性,在有限带宽条件下提高视频的主观视觉感知质量. A rate control method for SVC based on a spatiotemporal saliency map is presented in this paper. Mathe-matics models are introduced to describe the tasks in the human visual sysytem （HVS） mechanism, such as salient areas extraction and attention focus shift, and a spatiotemporal saliency map is obtained as a result. According to this map a rate allocation method based on greedy algorithm for enhancement layers is proposed to extract the substream, and the algorithm is optimized according to the characteristics of relation between human fovea and visual clarity. Experimentsshow that the proposed methods can simulate the characteristics of HVS and assure subjective visual quality when the bandwidth is limited.

关 键 词： 模式识别 视频编码 数据处理 时空显著图 码流提取 主观感知

领　　域： [自动化与计算机技术] [自动化与计算机技术]