Author Affiliations
1School of Intelligent Manufacturing, City College of Dongguan University of Technology, Dongguan, Guangdong 523419, China2School of Mechanical & Automotive Engineering, South China University of Technology, Guangzhou, Guangdong 510640, China;show less
Fig. 1. Flowchart of the proposed algorithm
Fig. 2. Initial disparity map before and after the adaptive weighting improvement. (a) Before adaptive weight improvement; (b) after adaptive weight improvement
Fig. 3. Disparity maps of each stage. (a) ArtL original image; (b) initial disparity map obtained by fusion cost of Census transformation and MI and improved guided filter aggregation; (c) disparity map optimized by MRF; (d) final disparity map obtained after segmentation optimization and median filtering
Fig. 4. Influence of ρ on the disparity map. (a) Influence of ρ on all area of the disparity map; (b) influence of ρ on the non-occlusion area of the disparity map
Fig. 5. Influence of k on the disparity map. (a) Influence of k on the non-occlusion area of the disparity map; (b) influence of k on all area of the disparity map
Fig. 6. Influence of k on the running time of each optimization stage
Fig. 7. Initial disparity map and optimized disparity map. (a) Initial disparity map; (b) color initial disparity map; (c) disparity map optimized by the proposed algorithm; (d) color disparity map optimized by the proposed algorithm
Parameter | λcen | λ1 | λ2 | Θ | ηΛ | ε | γ | a | κ | δ |
---|
Value | 25 | 0.3 | 0.7 | 2 | 3 | 10-5 | 20 | 2 | 16 | 0.01 |
|
Table 1. Experimental related parameter settings
k | 20 | 25 | 30 | 35 | 40 | 45 | 50 | 55 | 60 | 65 | 70 | 75 | 80 | 85 | 90 |
---|
Time /s | 19.8 | 8.9 | 7.4 | 6.8 | 6.4 | 5.8 | 5.5 | 5.0 | 5.3 | 4.5 | 4.7 | 4.2 | 4.3 | 4.3 | 4.2 |
|
Table 2. Influence of k on the average processing time of the images
Image | SED[27] | r200high[28] | ICSG[29] | LS_ELAS[30] | AD_Cens[31] | ADSM[32] | Proposed algorithm |
---|
Adirondack | 23.70 | 22.60 | 24.00 | 8.46 | 8.81 | 13.30 | 4.17 | ArtL | 15.60 | 12.80 | 6.93 | 3.83 | 5.27 | 6.10 | 3.57 | Jadeplant | 10.60 | 45.90 | 54.20 | 41.10 | 21.30 | 15.00 | 11.16 | Motorcycle | 18.30 | 12.00 | 12.00 | 5.12 | 4.10 | 3.67 | 2.45 | MotorcycleE | 17.70 | 12.00 | 10.40 | 5.80 | 5.58 | 5.67 | 2.36 | Piano | 17.70 | 19.20 | 15.60 | 5.54 | 6.57 | 7.08 | 8.82 | PianoL | 29.70 | 31.80 | 29.30 | 8.97 | 31.40 | 20.60 | 17.76 | Pipes | 28.50 | 15.30 | 18.40 | 7.44 | 8.50 | 6.57 | 3.92 | Playroom | 21.30 | 28.30 | 24.70 | 8.76 | 7.44 | 13.20 | 6.00 | Playtable | 18.20 | 36.40 | 26.70 | 22.40 | 24.80 | 23.10 | 5.87 | PlaytableP | 15.90 | 18.00 | 10.70 | 3.47 | 5.48 | 3.55 | 3.00 | Recycle | 16.20 | 18.30 | 17.70 | 6.93 | 3.32 | 5.76 | 4.11 | Shelves | 14.40 | 27.80 | 23.60 | 8.26 | 12.20 | 17.20 | 17.21 | Teddy | 6.65 | 10.70 | 7.73 | 2.29 | 3.11 | 3.05 | 1.95 | Vintage | 31.60 | 78.90 | 72.90 | 13.10 | 10.20 | 10.10 | 21.75 | Average error | 25.90 | 23.10 | 21.30 | 9.66 | 9.21 | 8.95 | 7.60 |
|
Table 3. Disparity average error of non-occluded pixel points obtained by different algorithms unit: %