Blind Stereo Image Quality Evaluation Based on Convolutional Network and Saliency Weighting

Table 3

Comparison of the results of the proposed method and several other methods in terms of the PLCC coefficient of performance.

Databases	Distortion	FR			Blind
		SSIM	FSIM	GMSD	Chen	Bensalma	Shao	Proposed
LIVE 3D Phase I	JP2K	0.868	0.937	0.928	0.855	0.848	0.872	0.908
	JPEG	0.496	0.601	0.652	0.476	0.376	0.597	0.622
	WN	0.938	0.931	0.947	0.9533	0.914	0.916	0.941
	Gblur	0.912	0.933	0.938	0.939	0.916	0.923	0.915
	ALL	0.899	0.936	0.943	0.929	0.895	0.899	0.913
LIVE 3D Phase II-symmetric	JP2K	0.8162	0.8183	0.875	0.670	0.690	0.903	0.898
	JPEG	0.6770	0.8456	0.844	0.601	0.551	0.873	0.890
	WN	0.9749	0.9630	0.961	0.946	0.936	0.917	0.982
	GBLUR	0.8325	0.8638	0.928	0.918	0.953	0.977	0.953
	ALL	0.7326	0.8301	0.925	0.814	0.823	0.912	0.918
LIVE 3D Phase II-asymmetric	JP2K	0.676	0.785	0.868	0.722	0.619	0.789	0.876
	JPEG	0.685	0.796	0.869	0.564	0.631	0.705	0.699
	WN	0.823	0.941	0.916	0.945	0.933	0.924	0.965
	GBLUR	0.840	0.888	0.741	0.692	0.862	0.855	0.951
	ALL	0.750	0.678	0.653	0.634	0.743	0.565	0.766