目的 由于色彩空间包含了图像的大量信息,而且Lab色彩空间更接近于人眼视觉,因此提出一种改进的无参考图像质量评价算法IQALE(image quality assessment using Lab color space and entropy),通过在SSEQ(spatial-spectral entropy-based quality)算法中加入Lab色彩空间a通道和b通道的特征来提高算法精度。方法 信息熵是近几年研究较多的图像特征,并且能较好地运用在图像质量评价研究中。该文在色彩空间和灰度空间同时提取信息熵特征,通过支持向量机(SVM)对图像特征和MOS值进行训练和测试。结果 在LIVE、TID2008、MICT、CSIQ和IVC这5个常用数据库上的实验结果表明:在算法中加入Lab色彩空间信息可以提高算法精度,并且本文算法IQALE的效果优于目前流行的无参考图像质量评价算法。为了验证算法的可扩展性,该文还在这5个数据库上进行了数据库独立性实验。结论 从实验结果来看,本文提出的IQALE算法通过加入色彩熵特征使得算法具有较高且较稳定的精度,数据库独立性实验也体现了算法较好的鲁棒性,对于各种失真类型都具有较好的普适性。

Objective An improved no-reference image quality assessment metric IQALE is proposed in this paper. Color space also includes lots of image information, and Lab color space is closer to human vision system. Therefore, to improve the metric accuracy, this paper adds a channel and b channel of Lab color space into the spatial-spectral entropy-based quality (SSEQ) algorithm. Method Information entropy is an image feature that is studied more in recent years, and can be applied to image quality assessment better. Information entropy is extracted in both color and gray spaces. Then the image features and MOS value are trained and tested via support vector machine (SVM). Result The results on LIVE, TID2008, MICT, CSIQ and IVC databases demonstrate that adding the information of Lab color space can improve the metric accuracy and IQALE algorithm is better than the recent popular no-reference image quality assessment algorithms. Moreover, in order to test the scalability of the proposed metric, the database independence experiment is conducted on the five image databases. Conclusion According to the result, IQALE method has better and more stable accuracy by adding the feature of color entropy. The database independence experiment also shows the better Robustness of the method. Furthermore, IQALE has better universality for every distortion type.