基于增强型无参考质量评价的序列图像自动选优策略*

针对序列图像，提出了增强型无参考质量评价的图像自动选优策略。首先在详细介绍SSEQ、NIQE和BIQI三种经典的无参考图像质量评价方法优缺点的基础上，提出了加权的质量评价策略以对序列图像进行自动选优。其次为了加快权重寻优的过程，提出了基于粒子群优化的PSO-WNRIQA算法。最后为了评估算法的性能，提出失序数比例DNR和失序对数DCNR作为算法评价标准。通过对LIVE Release2图像库中的实验结果表明，在jp2k失真、jpeg失真、高斯模糊失真和快速瑞利衰减失真图像的自动评优过程中，本文的策略相对SSEQ、NIQE和BIQI具有更优的性能，评价的准确性更高。能够应用于较大规模序列图像的自动质量评价。

Sequence images automatic selection strategy based on enhanced no-reference quality assessment

This paper proposed an enhanced no-reference quality assessment strategy for sequence images automatic selection. Firstly, on the basis of analyzing the advantages and disadvantages of three classic no-reference quality assessment methods, such as SSEQ, NIQE and BIQI, this paper proposed a weighted quality assessment strategy for sequence images automatic selection. Secondly, it proposed a PSO-WNRIQA algorithm based on particle swarm optimization to speed up the process of weighted selection. Finally, it proposed the disordered number ratio (DNR) and disordered couple number ratio (DCNR) as evaluation criteria to assess the performance of algorithms. The experimental results based on the LIVE Release2 images database show that the proposed strategy outperforms SSEQ, NIQE and BIQI with high accuracy for JPEG2000, JPEG, Gaussian blur and fast-fading distortions images. The proposed strategy is more suitable for large-scale sequence images quality assessment.