基于人眼视觉系统的误差分布反馈量化策略

Error Distribution Feedback Quantization Scheme Based on HVS

为了提高视频运动编码的质量,满足人眼视觉特性,降低量化编码失真,提出一种基于人眼视觉系统的误差分布反馈量化策略。该算法利用视频图像的运动目标与时间活动性关系,实现了运动场矢量分层描述,并通过各种活动性因子调节实际量化参数的大小。最后利用误差分布自适应的调整量化步长。实验结果表明,本文算法在不引入更多计算复杂度的同时,相同码率下的单位像素重建量化误差与图像质量(PSNR)优于TM5算法。

In order to improve the efficiency and quality of video motion coding, satisfy the properties of human visual, and reduce quantization distortion, an Error Distribution Feedback Quantization Scheme Based on HVS is proposed in this paper. Motion field layering is implemented by using the relationship between motion objects and temporal activity. And then the actual quantization parameter is adjusted by weighted coefficients which are the spatial and temporal activity factors. Finally this scheme utilizes error distri- bution to adjust quantization step adaptively. The experimental results show that the image that is reconstructed by novel scheme not only improving the lowest complexity, but also is better than the pictures quality coded TM5 quantization algorithm in errors per pixel and PSNR in the same bit rates.