论文部分内容阅读
根据人眼视觉的特点,提出一种基于视觉显著性分析的视频转码算法.首先利用H.264码流中的运动矢量场进行全局运动估计和局部运动分割,然后结合编码比特数的分布特点生成最终的视觉显著性图,最后在转码的重编码过程中,利用视觉显著性图对非显著性区域进行自适应频率系数压制,以在保持视频主观质量的前提下进一步提高压缩效率.实验结果表明:与现有视觉显著性区域检测方法相比,本算法能更准确完整地检测出视频中的显著性区域;而与参考的标准转码算法相比,平均降低约14%的编码比特率,同时保持了几乎相同的主观质量.
According to the characteristics of human vision, a video transcoding algorithm based on visual saliency analysis is proposed.Firstly, global motion estimation and local motion segmentation are performed by using the motion vector field in H.264 code stream. Combined with the distribution of the number of bits Finally, the final visual saliency map is generated. Finally, the visual saliency map is used to suppress the adaptive frequency coefficients in the non-salient regions to further improve the compression efficiency while maintaining the subjective quality of the video in the transcoding re-encoding process. The results show that compared with the existing visual saliency region detection method, the proposed algorithm can detect the salient regions in video more accurately and completely. Compared with the reference standard transcoding algorithm, the algorithm can reduce the average bit rate by about 14% Rate, while maintaining almost the same subjective quality.