Fig 1.
Block diagram of the DIBR system.
Fig 2.
Parallel camera configuration used for the generation of virtual stereoscopic images.
Fig 3.
Framework of the proposed structure-aided depth map smoothing approach.
Table 1.
Parameter presets for the domain transform filter used in testing.
Fig 4.
Comparison of smoothed depth maps using different all blur solutions.
A: original texture image, B: original depth map, C: symmetric filter, D: asymmetric filter, E: proposed filter with constraint ct1(u), F: proposed filter with constraints ct1(u) and ct2(u).
Fig 5.
Experimental results of different depth map fusion solutions.
A: detected hole regions in a right view, B: detected hole regions in a left view, C and F: influence map and processed depth map using our proposed method for a right view, D and G: influence map and processed depth map using our proposed method for a left view, E and H: influence map and processed depth map with method [20] for a left view.
Fig 6.
Processing flow for smoothing severely inaccurate depth maps with the proposed filter.
Fig 7.
Experimental results of optimized depth maps with our proposed filter.
A and E: original images, B: initial depth map of [25] (flower set) / F: initial depth map of [26] (castle set), C and G: refined depth maps, D and H: final optimized depth maps.
Fig 8.
Virtual view images of Interview (720 × 576 pixels).
A: no preprocessing, B: symmetric filter, C: asymmetric filter, D: distance dependent filter, E: proposed filter 1, F: proposed filter 2.
Fig 9.
Virtual view images of Ballet (1024 × 768 pixels).
A: no preprocessing, B: symmetric filter, C: asymmetric filter, D: distance dependent filter, E: proposed filter 1, F: proposed filter 2.
Table 2.
PSNR and SSIM comparison.
Table 3.
Computation time (s) comparison.
Table 4.
Results of subjective quality evaluation.
Fig 10.
Synthesized anaglyph images of test sets.
A: Interview, B: Ballet, C: Flower, D: Castle.