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

Table 4

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

Databases	Distortion	FR			Blind
		SSIM	FSIM	GMAD	Chen	Bensalm	Shao	Proposed
LIVE 3D Phase I	JP2K	0.867	0.901	0.905	0.871	0.817	0.900	0.881
	JPEG	0.456	0.563	0.610	0.435	0.328	0.607	0.594
	WN	0.938	0.930	0.947	0.939	0.906	0.927	0.931
	Gblur	0.899	0.925	0.937	0.921	0.918	0.924	0.858
	ALL	0.882	0.913	0.922	0.882	0.840	0.894	0.887
LIVE 3D Phase II-Symmetric	JP2K	0.726	0.824	0.867	0.662	0.608	0.904	0.872
	JPEG	0.718	0.841	0.838	0.630	0.548	0.910	0.884
	WN	0.945	0.937	0.927	0.907	0.924	0.937	0.942
	GBLUR	0.770	0.850	0.836	0.845	0.846	0.911	0.652
	ALL	0.700	0.909	0.910	0.837	0.805	0.897	0.893
LIVE 3D Phase II-Asymmetric	JP2K	0.724	0.805	0.854	0.722	0.619	0.789	0.872
	JPEG	0.714	0.805	0.876	0.636	0.678	0.696	0.686
	WN	0.882	0.952	0.937	0.929	0.941	0.924	0.955
	GBLUR	0.807	0.850	0.888	0.691	0.840	0.803	0.838
	ALL	0.719	0.661	0.642	0.611	0.697	0.524	0.711