【摘要】 近年来,基于结构相似性的视频质量的客观评价方法得到了广泛的关注,它建立在人眼对结构失真较为敏感这一假设之上,实验表明其性能优于峰值信噪比。该方法并没有考虑人眼视觉系统对不同的频率分量和对不同视频区域的敏感程度不同。基于人眼视觉系统的特点和结构相似性,提出了一种改进的视频质量评价方法,实验表明它与主观视频质量的相关性明显好于峰值信噪比和结构相似性。更多还原

【Abstract】 The video metric of structural similarity (SSIM) could achieve better performance than PSNR. Experiments show that the same attention paid to different frequency components in a frame could cause the relatively poor performance of SSIM. An improved video quality assessment based on HVS and SSIM (HSSIM) is proposed, in which sensitivity to various frequency components and unequal attentions to different regions of video frames are applied. Experiments indicate that this assessment method could achieve a higher correlation with perceived video quality than the original SSIM.更多还原