Method for Video Quality Evaluation Based on HVS Properties

摘要

摘要：本文针对视频质量的评测应用，对传统峰值信噪比(PSNR)算法加以改进。通过在视频帧内图像和帧间图像的处理过程中引入人眼视觉系统(HVS)的主要特性，克服传统 PSNR 算法在序列质量检测应用方面的缺陷。方法在帧内图像处理上利用人眼对边缘轮廓失真具有较强敏感性的特点，设计了基于图像边缘的检测方案以提高对典型空域失真的检测性能；在帧间图像处理上，通过测量帧间时域能量的变化，获得序列在时域轴上的典型特征，并据此对空域检测结果进行修正。通过上述改进，算法能在保持传统 PSNR 算法简易性的同时，提升其检测结果与主观感受的相关性；同时算法的计算量并不复杂，易于在检测设备中实现系统集成 Abstract: A method for the objective video quality assessment is proposed. By using the Human Vision System (HVS) properties into the intra and inter frame processing, the proposed method can overcome the traditional Peak Signal-to-noise Ratio (PSNR) in the application of video quality evaluation. Based on the principle that human's eyes tend to be more sensible to the degradation happening in the edge of image, a detection scheme is designed to evaluate the distortion in the intra frame. By measuring the temporal energy, know the typical features in the temporal domain and then use them to modify the evaluation result of the spatial domain. The proposed method inherits the advantages of the traditional PSNR such as simpleness and convenience and has better correlation with the subjective perception. And it can be easily implemented and integrated in the equipment system of detection.