论文部分内容阅读
在H.264编码过程中,帧间预测的最佳匹配块的选择和编码模式的判决由率失真代价函数决定,在该函数中,通常对失真采用的衡量方法是绝对(变换)误差和(SATD)。相应地,其他类似的衡量方法,如MSE和PSNR也被用在质量评估当中。然而,SA(T)D和PSNR已经被证明不能反映人眼视觉对失真的真实的敏感程度。最近,人们提出了一种称为HSSIM(基于人眼视觉系统的结构相似性)的新的图像质量测度,由于更好地考虑了图像的结构信息,HSSIM与人类视觉的一致性优于PSNR和MSE。文中提出了一种用于帧间编码的新的运动预测方法(MEHSSIM),它建立在基于人眼视觉系统的结构相似性的基础之上。实验表明,在由HSSIM测量的主观视频质量基本保持不变的情况下,新方法可以使码率降低平均13.8%。
In the H.264 encoding process, the choice of the best match block for inter prediction and the decision of the encoding mode are determined by the rate-distortion cost function, where the usual measure of distortion is absolute (transform) error and ( SATD). Accordingly, other similar measures, such as MSE and PSNR, are also used in the quality assessment. However, SA (T) D and PSNR have been shown not to reflect the real sensitivity of human vision to distortions. Recently, a new measure of image quality called HSSIM (based on the structural similarity of the human visual system) has been proposed. Due to the better consideration of the structural information of the image, the consistency between HSSIM and human vision is superior to PSNR and MSE. In this paper, a new motion prediction method (MEHSSIM) for inter-frame coding is proposed, which is based on the structural similarity of human visual system. Experiments show that the new method can reduce the bit rate by an average of 13.8% when the subjective video quality measured by HSSIM remains basically unchanged.