基于人眼视觉特性的自适应支持权重立体匹配算法 
下载： 1204次
Adaptive Support Weight Stereo Matching Algorithm Based on Human Visual Characteristics
南京航空航天大学机电学院, 江苏 南京 210016
图 & 表
图 1. (a)高斯分布式权重分配曲线;(b)原算法的权重分配曲线
Fig. 1. (a) Gaussian weight distribution curve; (b) weight distribution curve of original algorithm
下载图片 查看原文
图 2. HSI颜色模型
Fig. 2. HSI color model
下载图片 查看原文
图 3. 误匹配率分别随(a) σ和(b) λ变化情况
Fig. 3. Matching error rate versus (a) σ or (b) λ values
下载图片 查看原文
图 4. 视差计算结果。(a)真实视差图;(b)原ASW算法;(c)本文算法
Fig. 4. Results of parallax calculation. (a) Real parallax; (b) original ASW algorithm; (c) proposed algorithm
下载图片 查看原文
表 1分别只改进一项权重分配法则的实验结果
Table1. Experimental result with only one weight distribution rule improved%
| Condition | Venus | Cones | Teddy | Avg |
|---|
| Non | All | | Disc | Non | All | Disc | Non | All | Disc |
|---|
| Gauss only | 0.58 | 0.87 | 4.12 | 3.45 | 9.53 | 8.21 | 7.68 | 12.7 | 17.1 | 7.14 | | HSI only | 0.66 | 1.05 | 5.26 | 3.74 | 9.77 | 8.18 | 7.62 | 12.5 | 17.3 | 7.34 | | ASW | 0.71 | 1.19 | 6.13 | 3.97 | 9.79 | 8.26 | 7.88 | 13.3 | 18.6 | 7.76 |
|
查看原文
表 2本文算法和其他算法的误匹配率比较
Table2. Comparison of matching error rate with proposed algorithm and other algorithms%
| Algorithm | Venus | Cones | Teddy | Avg |
|---|
| Non | All | | Disc | Non | All | Disc | Non | All | Disc |
|---|
| Region TreeDP[13] | 0.22 | 0.57 | 1.93 | 6.31 | 11.9 | 11.8 | 7.42 | 11.9 | 16.8 | 7.65 | | DoubleBP[14] | 0.31 | 0.45 | 1.87 | 2.90 | 8.78 | 7.79 | 3.53 | 8.30 | 9.63 | 4.84 | | HEBF[15] | 0.22 | 0.33 | 2.41 | 2.78 | 9.28 | 8.10 | 6.54 | 11.8 | 15.2 | 6.29 | | Cross-Based[16] | 0.62 | 0.96 | 3.20 | 6.28 | 12.7 | 12.9 | 9.75 | 15.1 | 18.2 | 8.86 | | ASW | 0.71 | 1.19 | 6.13 | 3.97 | 9.79 | 8.26 | 7.88 | 13.3 | 18.6 | 7.76 | | Proposed | 0.54 | 0.82 | 3.81 | 3.37 | 9.43 | 8.12 | 7.49 | 12.6 | 16.1 | 6.92 |
|
查看原文
表 3本文算法和其他算法的误匹配率和运算时间比较
Table3. Comparison of matching error rate and operation time with proposed algorithm and other algorithms
| Algorithm | Venus | Cones | Teddy | Avg /% |
|---|
| Non /% | All /% | | Disc /% | Time /s | Non /% | All /% | Disc /% | Time /s | Non /% | All /% | Disc /% | Time /s |
|---|
| CAG | 1.58 | 2.02 | 9.90 | 8.59 | 4.98 | 10.8 | 12.3 | 10.5 | 10.1 | 15.7 | 21.9 | 9.31 | 9.92 | | TP | 0.34 | 0.92 | 4.55 | 10.8 | 3.10 | 9.31 | 8.59 | 14.5 | 9.83 | 15.3 | 20.3 | 14.9 | 8.03 | | FBS | 0.40 | 0.81 | 3.38 | 15.6 | 3.48 | 9.34 | 8.81 | 20.1 | 7.65 | 13.3 | 16.2 | 21.3 | 7.04 | | Proposed | 0.54 | 0.82 | 3.81 | 38.2 | 3.37 | 9.43 | 8.12 | 49.7 | 7.49 | 12.6 | 16.1 | 51.3 | 6.92 |
|
查看原文
刘雪松, 沈建新, 张燕平. 基于人眼视觉特性的自适应支持权重立体匹配算法[J]. 激光与光电子学进展, 2018, 55(3): 031013. Xuesong Liu, Jianxin Shen, Yanping Zhang. Adaptive Support Weight Stereo Matching Algorithm Based on Human Visual Characteristics[J]. Laser & Optoelectronics Progress, 2018, 55(3): 031013.
PDF全文
