Robust semi-automatic 2D-to-3D image conversion via residual-driven optimization

EURASIP Journal on Image and Video Processing

Table 1 SSIM of estimated depth on RGBZ datasets when cross-boundary scribbles are present

	RW	HGR	NRW	OPT	OCP	SDF	ℓ ₁	Proposed
RGBZ _ 01	0.836	0.776	0.852	0.842	0.873	0.839	0.894	0.932
RGBZ _02	0.817	0.849	0.827	0.821	0.821	0.817	0.878	0.931
RGBZ _03	0.853	0.868	0.868	0.867	0.861	0.855	0.808	0.890
RGBZ _04	0.911	0.911	0.923	0.928	0.897	0.917	0.808	0.944
RGBZ _05	0.901	0.904	0.916	0.906	0.911	0.903	0.935	0.929
RGBZ _06	0.907	0.885	0.906	0.879	0.880	0.905	0.909	0.921
RGBZ _07	0.895	0.872	0.889	0.886	0.906	0.892	0.904	0.912
RGBZ _08	0.933	0.906	0.934	0.933	0.944	0.932	0.934	0.951
RGBZ _09	0.927	0.750	0.938	0.916	0.926	0.925	0.900	0.950
Average	0.887	0.858	0.895	0.886	0.891	0.887	0.885	0.929

The first and second best SSIM at each row are shown in bold and italics, respectively