目前图像质量评价的研究都是在相对标准观察环境下进行的, 没有考虑电子显示设备在不同场合使用时照明亮度的变化对图像失真造成的影响。通过在四种不同环境光照度下进行主观评价实验探究MOS值变化规律, 在BRISQUE和SSEQ算法的研究基础上, 利用机器学习方式获取亮度因子来反映不同亮度条件引起的人眼感知变化, 建立不同亮度下图像质量评价模型。改进后的算法模型针对噪声、压缩、模糊的SROCC系数在不同的亮度环境下均有良好的效果, 表明修正后的算法模型具有很高的普适性和准确性。所提算法是对现有算法模型的很好补充, 尤其适用于户外高亮度条件下的图像质量评价。

The current research on image quality assessment is carried out in a relative standard observation environment, but the influence of the illumination on image distortion is not considered when the electronic display device is used in different operating environments. Based upon BRISQUE and SSEQ algorithms, an illumination factor is established by machine training to adapt to the changes in human perception caused by different ambient illumination. The MOS values are obtained by a subjective evaluation experiment using four different ambient illuminations. Based upon these data, a new image quality assessment model is proposed that would apply to different ambient illumination. The improved algorithm performs well in evaluating noise, compression and blurred images under different illumination, which indicates that the modified algorithm has high universality and accuracy. This new algorithm complements the existing no-reference image quality assessment model, especially under high ambient illumination.