基于失真类型预测的图像质量评价方法

目的克服单个图像质量评价方法针对各种失真类型的性能表现不均衡的问题。方法首先从多个角度测量参考与测试图像的视觉相似性(包括人眼视觉系统,低层次视觉处理和本地纹理),接着分别利用粗糙集和支持向量机完成特征简化与图像失真类型预测,并最终根据测试图像所遭受的失真类型选取适当的质量评价算法。结果提出一种基于失真类型预测的图像质量评价算法。结论该算法在整体性能上要优于传统的图像质量评价算法,更好地体现了主客观评价的一致性;针对单个失真类型开发性能优的质量评价方法,可以进一步提升该算法的性能。

Image quality assessment based on the distortion type predication

Aim To overcome the performance imbalance of individual image quality assessment algorithms with respect to various distortion types.Methods First the visual similarity is measured between reference and test images from several perspectives(including the human vision system,the low-level visual processing and local textures),then the dimension of features is reduced and the distortion type is predicted via rough set and Support Vector Machine respectively,finally the proper image quality assessment method is chosen according to the distortion type that the test image suffers.Results An image quality assessment method based on the distortion type predication is proposed.Conclusion The proposed method outperforms the traditional image quality assessment algorithms,by agreeing with subjective ratings in a better way;developing image quality assessment methods for single distortion types with a better performance than current methods could promote this method further.