基于内容划分和运动补偿的视频质量评价

目前基于结构相似性的图像质量评价算法均是对图像进行整体质量分析,但人类视觉系统对图像中不同部分的敏感程度不同,而对图像进行整体质量分析无法有效反映出这些差异。鉴于此,提出了一种基于内容划分的图像质量评价算法,根据图像不同区域的梯度将图像分为四个部分,分别进行质量评价。之后,采用基于运动补偿的帧加权方式将上述方法扩展为视频质量评价。实验证明,所述算法与目前比较流行的几个算法相比具有较高的评价准确性。

Video quality assessment based on content-partitioned approach and motion compensation

Current structural similarity based image quality assessment algorithm is generally the overall image quality analysis. However, for human visual system, different regions in image have different visual sensitivities, and the current method can not reflect these differences effectively. In this view, this paper proposed a content-partitioned image quality assessment algorithm, which partitioned an image into four regions according to their different gradient magnitudes and assessed their qualities respectively. Then, based on motion compensation, this paper proposed a frame-weighting approach, which extended the proposed algorithm to video quality assessment. The experiments show that the proposed algorithm is more accurate than several current popular algorithms.