Dense Disparity Map Extraction Method Based on Improved Convolutional Neural Network

Dongzhen Huang; Qin Zhao; Huawei Liu; Baoqing Li; Xiaobing Yuan

doi:10.3788/LOP55.121004

Laser & Optoelectronics Progress, Volume. 55, Issue 12, 121004(2018)

Dense Disparity Map Extraction Method Based on Improved Convolutional Neural Network

Dongzhen Huang^1,2, Qin Zhao^1,2, Huawei Liu¹, Baoqing Li¹, and Xiaobing Yuan^1、*

¹ Key Laboratory of Microsystem Technology, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 201800, China

² University of Chinese Academy of Sciences, Beijing 100049, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(10)

Fig. 1. Dual pyramid network structure

Download full size

Fig. 2. Comparison of network structures

Download full size

Fig. 3. Relationship between error rate and parameters of additional convolutional layers. (a) Error rate with different convolutional kernel sizes; (b) test error with different number of convolutional layers

Download full size

Fig. 4. Comparison of disparity maps. (a) Signpost 1; (b) fence; (c) car; (d) signpost 2; (e) traffic sign; (f) signpost 3

Download full size

Fig. 5. Zoomed parts over region marked by box in Fig. 4. (a) Signpost 1; (b) fence; (c) car; (d) signpost 2; (e) traffic sign; (f) signpost 3

Download full size

Fig. 6. Effect of structure on subjective quality. (a) Left input image; (b) without additional convolutional layers; (c) without dual pyramid structure; (d) default

Download full size

Table 1. Comparison of complexity and accuracy
View table
Table 1. Comparison of complexity and accuracy
Item MC-CNN-fst Proposed network
3×3 5×5
Time consumption /s 0.1902 0.2561 0.2438
Test error /% 3.029 2.784 2.795

Table 2. Objective index of 10 test images
View table
Table 2. Objective index of 10 test images
Image MC-CNN-fst Proposed (5×5)
1 0.807 0.807
2 4.180 3.892
3 1.210 1.147
4 1.207 1.117
5 8.960 8.006
6 6.228 5.866
7 1.588 1.475
8 0.293 0.213
9 2.934 2.658
10 3.020 2.877
Average 3.029 2.795

Table 3. Comparison of complexity
View table
Table 3. Comparison of complexity
Item MC-CNN-fst Proposed(5×5)
Complexity ofconvolution layers 111168 221760
Time consumption /s 0.1902 0.2438
Whole timeconsumption /s 0.4040 0.4770

Table 4. Comparison of different network structures
View table
Table 4. Comparison of different network structures
Item MC-CNN-fst MC-CNN-fst(1×1) Without additionalconvolutional layers Without dualpyramid structure Default
3 pixel /% 3.029 3.017 3.045 2.757 2.795
Time consumption /s 0.1902 0.2081 0.2251 0.2252 0.2438

Tools

Get Citation

Copy Citation Text

Dongzhen Huang, Qin Zhao, Huawei Liu, Baoqing Li, Xiaobing Yuan. Dense Disparity Map Extraction Method Based on Improved Convolutional Neural Network[J]. Laser & Optoelectronics Progress, 2018, 55(12): 121004

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites