王茜,郑斌军,孔玲君,顾萍.基于视觉显著性和感知相似性的全参考图像质量评价方法[J].包装工程,2022,43(9):239-248.
WANG Qian,ZHENG Bin-jun,KONG Ling-jun,GU Ping.Full Reference Image Quality Assessment Based on Visual Saliency and Perception Similarity Index[J].Packaging Engineering,2022,43(9):239-248. |
| 基于视觉显著性和感知相似性的全参考图像质量评价方法 |
| Full Reference Image Quality Assessment Based on Visual Saliency and Perception Similarity Index |
| |
| DOI：10.19554/j.cnki.1001-3563.2022.09.032 |
| 中文关键词: 图像质量评价 视觉显著性 人类视觉系统 |
| 英文关键词: image quality assessment visual saliency human visual system |
| 基金项目:一流专科高等职业教育专业建设项目(2020ylxm–1) |
|
| 摘要点击次数: |
| 全文下载次数: |
| 中文摘要: |
| 目的 图像质量评价(IQA)旨在使用计算模型自动衡量和评价图像质量，以代替人类视觉系统的主观意见，并应用到相关实际问题中。方法 首先将参考图像与失真图像进行输入，使用视觉显著性模型计算图像局部相似度的特征映射，并在质量得分池化阶段作为加权函数，同时，针对视觉显著性图作为单一特征映射的不足，增加了梯度幅度，然后将图像进行颜色空间的转化提取颜色特征，最后分配相应的权重来计算图像相似度。结果 在4个大型数据集上的对比测试显示，在保持适度计算复杂度的同时，VSPSI相比其他有代表性的模型在预测精度上得到了一定的提升，特别是在TID2013数据集上的SROCC达到了0.905 5。结论 研究结果表明，VSPSI是一个性能优良的IQA方法，在不同数据集和不同失真类型中都有良好的表现，具有较强的鲁棒性，可胜任多类失真图像的客观质量评价，同时可通过优化视觉显著性模型进一步提升VSPSI的性能。 |
| 英文摘要: |
| Image Quality Assessment (IQA) is designed to use computational models to automatically measure image quality in line with the subjective assessment of the human visual system and to apply them to relevant practical problems. Firstly, the reference image and the distorted image are input, and the visual saliency model is used to calculate the feature mapping of the local similarity of the image, which is used as the weighting function in the quality score pooling stage. At the same time, in view of the deficiency of the visual saliency map as a single feature mapping, the gradient amplitude is increased, then the image is transformed into the color space to extract the color features, and finally the corresponding weight is allocated to calculate the image similarity. Results the comparative test on four large data sets shows that while maintaining a moderate computational complexity, vspsi has improved the prediction accuracy compared with other representative models. In particular, the SROCC on the tid2013 data set reaches 0.905 5. The results tell that VSPSI is an IQA with excellent performance. It has good performance in different data sets and different distortion types, and has strong robustness. It can be used to assess the objective quality of multi class distorted images. At the same time, the performance of VSPSI can be further improved by optimizing the visual saliency model. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
渝公网安备 50010702501716号