Acta Optica Sinica, Volume. 43, Issue 14, 1415002(2023)

Depth Estimation Using Polarizer-Free Liquid Crystal Lens

Wenjie Lai1, Zhiqiang Liu1, Tao Sun2, and Xiao Hu1、*
Author Affiliations
  • 1School of Optoelectronic Science and Engineering, University of Electronic Science and Technology of China, Chengdu 610000, Sichuan, China
  • 2Armored Forces Research Institute, Army Research Academy, Beijing 100072, China
  • show less
    Figures & Tables(15)
    Light path diagram of polarizer-free liquid crystal lens DFD model
    Numerical analysis of error indicator S. (a) Comparison of S distributions for natural light and extraordinary light in frequency domain; (b) relationship between error indicator S and fuzzy spot radius when v=0.8vmax and R is different
    Relationship between power of liquid crystal lens and voltage difference. (a) Interference image; (b) relationship between fitted power of liquid crystal lens (negative values mean negative lens) and voltage difference
    Relatively large errors in initial calculation results. (a) Near-focus image; (b) far-focus image; (c) estimated depth; (d) ground truth
    Foreground depth estimation affects background depth estimation. Dashed line is a reference line for convenience of comparison. (a) Near-focus image; (b) far-focus image; (c) foreground depth estimation affects background depth estimation (sliding window size is63×63); (d) as sliding window size increases (127×127), adverse effect of foreground on background intensifies
    Confidence after correction. (a) Far-focus image; (b) near-focus image; (c) estimated depth; (d) according to original confidence, retain 53% high-confidence data (dark data in image is excluded data); ( e) according to improved confidence, retain 53% high-confidence data; (f) ground truth; (g) depth completion of (d); (h) depth completion of (e)
    Effect comparison of depth completion methods. (a) Semantic segmentation + Laplacian matting; (b) Laplacian matting; (c) Markov random field; (d) ground truth
    Effect comparison of depth estimation without polarizer and with polarizer for slop scenes. First and second rows are slop scenes with polarizer and without polarizer, respectively. First and second columns are images of liquid crystal lens with optical power of -0.70 m-1 and 1.86 m-1, respectively. Third column is depth estimation result. Fourth column is ground truth
    Effect comparison of depth estimation without polarizer and with polarizer for plane scenes. First and second rows are plane scenes with polarizer and without polarizer, respectively. First and second columns are images of liquid crystal lens with optical power of 0 and 0.95 m-1, respectively. Third column is depth estimation result. Fourth column is ground truth
    Comparison of input images. From left to right are images formed by liquid crystal lens with optical power of -1, -0.7, and 1.86 m-1, respectively. First row are images with polarizer and second row are images without polarizer
    Optimal depth estimation for different depth ranges. Depth ranges corresponding to first and second rows are doll, nesting doll, plush toy, zebra, and background. Third and fourth rows correspond to true depth value, global optimal initial estimation value, and depth estimation values after error elimination and depth completion. First and third rows are results with polarizer. Second and fourth rows are results without polarizer
    • Table 1. Adjustment values of optical power of liquid crystal lens

      View table

      Table 1. Adjustment values of optical power of liquid crystal lens

      NumberPlc /m-1NumberPlc /m-1NumberPlc /m-1NumberPlc /m-1
      1-2.005-0.7090.64131.86
      2-1.656-0.37100.95
      3-1.3670111.24
      4-1.0080.31121.56
    • Table 2. Optimal depth estimation parameters for slope and plane scenes (P means with polarizer and NP means without polarizer in Type column)

      View table

      Table 2. Optimal depth estimation parameters for slope and plane scenes (P means with polarizer and NP means without polarizer in Type column)

      SceneTypedavg /mPlc1 /m-1Plc2 /m-1RMSE /mAWT1.25Re1Re2
      SlopP0.77-0.371.560.071.0012.839.41
      NP0.77-0.371.560.051.0012.839.41
      PlaneP0.900.310.640.021.002.781.03
      NP0.9000.950.021.006.354.60
    • Table 3. Depth ranges for different objects

      View table

      Table 3. Depth ranges for different objects

      ObjectDepth /m
      Doll0.38
      Nesting doll0.70
      Plush toy0.99
      Zebra1.27
      Background1.96
    • Table 4. Effect comparison of optimal depth estimation for different depth ranges under extraordinary light and natural light. First sub-row in each row represents extraordinary light, second sub-row in each row represents natural light, and bold indicates better effect

      View table

      Table 4. Effect comparison of optimal depth estimation for different depth ranges under extraordinary light and natural light. First sub-row in each row represents extraordinary light, second sub-row in each row represents natural light, and bold indicates better effect

      ObjectNdmin /mdmax /mdavg /mdstd /(10-3 m)Plc1 /m-1Plc2 /m-1Re1Re2RoRMSEAWT1.25
      Doll1182140.200.490.384.51.561.865.852.390.110.94
      01.8623.822.3923.820.240.62
      Nesting doll569190.490.780.707.40.640.952.590.990.031.00
      01.569.968.029.960.080.96
      Plushtoy260770.781.060.998.90.310.641.692.120.051.00
      0.310.641.692.125.260.051.00
      Zebra448701.061.351.2739.800.312.650.920.081.00
      -0.370.956.918.302.650.071.00
      Background673701.932.211.9610.4-0.370.313.744.100.110.99
      -0.370.313.744.100.530.120.98
      All3148180.202.500.96608.0-0.701.8613.6515.850.430.50
      -1.001.8617.1115.855.590.320.69
      All3148180.202.500.96608.0-0.701.8613.6515.850.400.51
      -1.001.8617.1115.855.590.190.87
    Tools

    Get Citation

    Copy Citation Text

    Wenjie Lai, Zhiqiang Liu, Tao Sun, Xiao Hu. Depth Estimation Using Polarizer-Free Liquid Crystal Lens[J]. Acta Optica Sinica, 2023, 43(14): 1415002

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Machine Vision

    Received: Feb. 16, 2023

    Accepted: Mar. 24, 2023

    Published Online: Jul. 13, 2023

    The Author Email: Xiao Hu (huxiao@uestc.edu.cn)

    DOI:10.3788/AOS230562

    Topics