论文部分内容阅读
图像质量的好坏是评价图像压缩算法优劣的重要依据．传统的图像质量客观评价方法不能反映出图像在视觉感知上的失真，而主观评价方法如平均主观分数（ＭＯＳ）带有很强的不确定性．建立在人眼视觉模型（ＨＶＳ）基础上的感知均方误差（ＰＭＳＥ）做为主客观联系的桥梁，能有效地反映出人对图像失真在视觉上的感知．类似于均方误差（ＭＳＥ）的定义，我们采用二范数测度的失真测量来定义ＰＭＳＥ，并用ＷｅｂｅｒＦｅｃｈｎｅｒ衰减、调制传递函数（ＭＴＦ）过程和Ｂ样条小波进行模拟．我们对感知误差信号进行误差分割，对不同类型的误差进行不同的处理．最后去相关，加权和，得到一客观值，用于图像质量的客观评价．实验表明ＰＭＳＥ和ＭＯＳ具有很强的相关性．
The quality of the image is an important basis to evaluate the advantages and disadvantages of the image compression algorithm. Traditional methods of objective evaluation of image quality can not reflect the visual perception of image distortion, and subjective evaluation methods such as the average subjective score (MOS) with a strong uncertainty. The perceived mean square error (PMSE) based on human visual system model (HVS) serves as a bridge between the subjective and the objective, which can effectively reflect the visual perception of human image distortions. Similar to the definition of mean square error (MSE), we define PMSE using distortion measures of two-norm measure and simulate with Weber-Fechner attenuation, modulation transfer function (MTF) process and B-spline wavelet. We error-sense the error signal segmentation, different types of errors for different treatment. Finally, to the relevant, weighted sum, get an objective value, for the objective evaluation of image quality. Experiments show that PMSE and MOS have a strong correlation.