A Feature-Enriched Completely Blind Image Quality Evaluator

Existing blind image quality assessment (BIQA) methods are mostly opinion-aware. They learn regression models from training images with associated human subjective scores to predict the perceptual quality of test images. Such opinion-aware methods, however, require a large amount of training samples with associated human subjective scores and of a variety of distortion types. The BIQA models learned by opinion-aware methods often have weak generalization capability, hereby limiting their usability in practice. By comparison, opinion-unaware methods do not need human subjective scores for training, and thus have greater potential for good generalization capability. Unfortunately, thus far no opinion-unaware BIQA method has shown consistently better quality prediction accuracy than opinion-aware methods. Here we aim to develop an opinion-unaware BIQA method that can compete with, and perhaps outperform existing opinion-aware methods. By integrating natural image statistics features derived from multiple cues, we learn a multivariate Gaussian model of image patches from a collection of pristine natural images. Using the learned multivariate Gaussian model, a Bhattacharyya-like distance is used to measure the quality of each image patch, then an overall quality score is obtained by average pooling. The proposed BIQA method does not need any distorted sample images nor subjective quality scores for training, yet extensive experiments demonstrate its superior quality-prediction performance to state-of-the-art opinion-aware BIQA methods.

The source code for the proposed blind NR-IQA metric IL-NIQE can be downloaded here: ILNIQE.zip.

In order to facilitate the other researchers, the prediction scores obtained by IL-NIQE along with MOS (or DMOS) on different datasets are provided here.

With the above mat file, the performance metrics could be easily computed. For example, with the following codes, the IL-NIQE's SROCC on TID2013 can be obtained :

matData = load('ILNIQEOnTID2013.mat');
ILNIQEOnTID2013= matData.ILNIQEOnTID2013;
SROCC = corr(ILNIQEOnTID2013(:,1), ILNIQEOnTID2013(:,2), 'type', 'spearman');

Four benchmark large-scale IQA datasets are used to evaluate the proposed IL-NIQE index, including TID2013, CSIQ, LIVE, and LIVE Multiply Distortion. The LIVE Multiply Distortion (MD) dataset was constructed by two sections and we regard them as two separate datasets, denoted by LIVE MD1 and LIVE MD2.

Since opinion-aware methods need to use the distorted images in the dataset to learn the model, we partition the dataset into a training subset and a testing subset. We report results under three partition settings: distorted images associated to 80%, 50%, and 10% of the reference images are used for training and the remaining for testing. The partition was randomly conducted 1,000 times and the median results are reported in the Table 1. For IL-NIQE, NIQE and QAC, though they do not need training on the dataset, we report their results on the partitioned test subset to make the comparison consistent.

Datasets	Methods	80%	50%	10%
Datasets	Methods	SRCC PLCC	SRCC PLCC	SRCC PLCC
TID2013	BIQI	0.349 0.366	0.332 0.332	0.199 0.250
	BRISQUE	0.573 0.651	0.563 0.645	0.513 0.587
	BLIINDS2	0.536 0.628	0.458 0.480	0.402 0.447
	DIIVINE	0.549 0.654	0.503 0.602	0.330 0.391
	CORNIA	0.549 0.613	0.573 0.652	0.508 0.603
	NIQE	0.317 0.426	0.317 0.420	0.313 0.398
	QAC	0.390 0.495	0.390 0.489	0.372 0.435
	IL-NIQE	0.521 0.648	0.513 0.641	0.494 0.590
CSIQ	BIQI	0.092 0.237	0.092 0.396	0.020 0.311
	BRISQUE	0.775 0.817	0.736 0.781	0.545 0.596
	BLIINDS2	0.780 0.832	0.749 0.806	0.628 0.688
	DIIVINE	0.757 0.795	0.652 0.716	0.441 0.492
	CORNIA	0.714 0.781	0.678 0.754	0.638 0.732
	NIQE	0.627 0.725	0.626 0.716	0.624 0.714
	QAC	0.486 0.654	0.494 0.706	0.490 0.707
	IL-NIQE	0.822 0.865	0.814 0.854	0.813 0.852
LIVE	BIQI	0.825 0.840	0.739 0.764	0.547 0.623
	BRISQUE	0.933 0.931	0.917 0.919	0.806 0.816
	BLIINDS2	0.924 0.927	0.901 0.901	0.836 0.834
	DIIVINE	0.884 0.893	0.858 0.866	0.695 0.701
	CORNIA	0.940 0.944	0.933 0.934	0.893 0.894
	NIQE	0.908 0.908	0.905 0.904	0.905 0.903
	QAC	0.874 0.868	0.869 0.864	0.866 0.860
	IL-NIQE	0.902 0.906	0.899 0.903	0.899 0.903
MD1	BIQI	0.769 0.831	0.580 0.663	0.159 0.457
	BRISQUE	0.887 0.921	0.851 0.873	0.829 0.860
	BLIINDS2	0.885 0.925	0.841 0.879	0.823 0.859
	DIIVINE	0.846 0.891	0.805 0.836	0.631 0.675
	CORNIA	0.904 0.931	0.878 0.905	0.855 0.889
	NIQE	0.909 0.942	0.883 0.921	0.874 0.912
	QAC	0.418 0.597	0.406 0.552	0.397 0.541
	IL-NIQE	0.911 0.930	0.899 0.916	0.893 0.907
MD2	BIQI	0.897 0.919	0.835 0.860	0.769 0.773
	BRISQUE	0.888 0.915	0.864 0.881	0.849 0.867
	BLIINDS2	0.893 0.910	0.852 0.874	0.850 0.868
	DIIVINE	0.888 0.916	0.855 0.880	0.832 0.851
	CORNIA	0.908 0.920	0.876 0.890	0.843 0.866
	NIQE	0.834 0.884	0.808 0.860	0.796 0.852
	QAC	0.501 0.718	0.480 0.689	0.473 0.678
	IL-NIQE	0.928 0.915	0.890 0.895	0.882 0.896

For the five opinion-aware BIQA methods, their quality prediction models trained on the entire LIVE dataset are provided by the original authors. Thus, we directly use them for testing on the other datasets. The results are shown in Table 2. For each performance measure, the two best results are highlighted in bold. In Table 3 we present the weighted-average SRCC and PLCC indices of all methods over the four datasets, and the weight assigned to each dataset linearly depends on the number of distorted images contained in that dataset. Furthermore, we train the opinion-aware methods on the entire TID2013 dataset and then perform testing on the rest datasets. The results are shown in Table 4 and Table 5.

	TID2013	CSIQ	MD1	MD2
	SRCC PLCC	SRCC PLCC	SRCC PLCC	SRCC PLCC
BIQI	0.394 0.468	0.619 0.695	0.654 0.774	0.490 0.766
BRISQUE	0.367 0.475	0.557 0.742	0.791 0.866	0.299 0.459
BLIINDS2	0.393 0.470	0.577 0.724	0.665 0.710	0.015 0.302
DIIVINE	0.355 0.545	0.596 0.697	0.708 0.767	0.602 0.702
CORNIA	0.429 0.575	0.663 0.764	0.839 0.871	0.841 0.864
NIQE	0.311 0.398	0.627 0.716	0.871 0.909	0.795 0.848
QAC	0.372 0.437	0.490 0.708	0.396 0.538	0.471 0.672
IL-NIQE	0.494 0.589	0.815 0.854	0.891 0.905	0.882 0.897

	LIVE	CSIQ	MD1	MD2
	SRCC PLCC	SRCC PLCC	SRCC PLCC	SRCC PLCC
BIQI	0.047 0.311	0.010 0.181	0.156 0.175	0.332 0.380
BRISQUE	0.088 0.108	0.639 0.728	0.625 0.807	0.184 0.591
BLIINDS2	0.076 0.089	0.456 0.527	0.507 0.690	0.032 0.222
DIIVINE	0.042 0.093	0.146 0.255	0.639 0.669	0.252 0.367
CORNIA	0.097 0.132	0.656 0.750	0.772 0.847	0.655 0.719
NIQE	0.906 0.904	0.627 0.716	0.871 0.909	0.795 0.848
QAC	0.868 0.863	0.490 0.708	0.396 0.538	0.471 0.672
IL-NIQE	0.898 0.903	0.815 0.854	0.891 0.905	0.882 0.897

	BIQI	BRISQUE	BLIINDS2	DIIVINE	CORNIA	NIQE	QAC	IL-NIQE
SRCC	0.458	0.424	0.424	0.435	0.519	0.429	0.402	0.599
PLCC	0.545	0.548	0.525	0.595	0.643	0.512	0.509	0.675