基于视觉敏感度的JPEG图像质量评价

JPEG Image Quality Assessment Based on Visual Sensitivity

为解决无参考图像质量评价与人眼视觉系统(HVS)特征的一致性问题,提出一种基于视觉敏感度的JPEG图像质量评价方法。采用支持向量回归神经网络学习和模拟HVS特征与平均主观得分之间的函数关系,利用边缘幅度和长度、背景活动度和亮度等视觉敏感度特征,实现符合HVS特征的无参考图像质量评价。实验结果表明,该方法的误差小、精度高、预测性能好,并与HVS感知特征具有高度一致性。

This paper presents a visual sensitivity measurement method to assess the visual quality of JPEG images for consistency between image quality assessment without reference image and Human Visual System（HVS） features.Support Vector Regression Neural Network（SVR-NN） is used to approximate the functional relationship between HVS feature and Mean Opinion Score（MOS）.The measuring of visual quality of JPEG images is realized using HVS features such as edge amplitude and length,background activity and luminance.Experimental results show that the method has less error,high accuracy,excellent estimation,and exhibits much higher correlation with HVS perception feature.