1. Abstract ID:2276 Programm Nr 2276
3D Quantitative Evaluation System for Integral Photography based 3D Autostereoscopic Medical Display
Zhencheng Fan, Sen Zhang, Yitong Weng, Guowen Chen and Hongen Liao
Department of Biomedical Engineering, School of Medicine,
Tsinghua University, Beijing, China
模板 更新成英文 清华大学重点突出一下

2. Background 3D medical display
--- 3D images with spatial information , assist surgeons
Main categories
Stereoscopic display
need supplementary instruments
fatigue
not for multi-viewers
Auto-stereoscopic display
without supplementary instruments
multi-viewers
a promising one
Google glass
Da Vinci 图像
IP based 3D display

3. Background 3D Quantitative Evaluation System Main parameters
Viewing angle
Depth of the 3D displayed object
Image quality
Demands:
Evaluation systemic, objective, quantitative
强调autostereoscopic displayer factors
现有的研究对参数并没有定量的三维评估，
背景（加入）
图像 删掉
存在的问题点 和 评估的参数
3D Quantitative Evaluation System

4. 3D auto-stereoscopic displayer
Methods
System Overview
stereo-camera
3D auto-stereoscopic displayer
workstation
platform
auto-control
image
processing
image acquisition
改图 相机方向 平台（不要凸显 或者是改小 或者是用箭头显示）

5. Methods Main factors to be evaluated
Image quality --maintenance of image details
-Modulation Transfer Function (MTF)
-2D correlation coefficient
Depth of the 3D displayed object
-Relative deviation of depth displayed (RDD)
Viewing angle
-the maximum angle without flipping

6. Methods Design of 3D evaluation reference patterns
3D zone plate pattern
To measure the image quality
Generation formula:
|x| is the radius of the displayed object
g(x) is the intensity of the generated pattern
1/8 3D zone plate pattern:
Pattern with double corners
to measure the viewing angle
to measure the RDD
Spatial relationship
20mm

7. Experiments IP based 3D autostereoscopic display Integral Photography
without using supplementary instruments
full parallax, multi-viewers
point retracing rendering algorithm
H. Liao, et al. (2000)
显示器 构造 性能 参数 用的方法
IP的原理

8. 3D autostereoscopic medical display
Experiments
IP based pattern
Evaluation system
3D autostereoscopic medical display
platform
stereo-camera
3D display
nexus10(300dpi)
lens (pitch:1.016mm, gap:3mm)
Rotation platform --rotation
XYZ platform --movement
Stereo-camera --(Guppy PRO F125c) *960

9. 3D autostereoscopic medical display
Experiments
Camera calibration
3D autostereoscopic medical display
platform
stereo-camera
10*7 chessboard
Zhang’s method
Viewing angle
One camera
Rotation: rotate angle 1°
Serious of images
Image with no flipping
Depth displayed
Stereo-camera
Computer vision
Image quality
One camera
Formula mentioned above

10. 2D correlation coefficient after logarithmic scaling
Results
Viewing angle
Parameters
1/°
2/°
3/°
Average/°
Error
horizontal
viewing angle
17.0
16.0
18.0
-11.6%
vertical
19.5
22.0
21.0
20.8
-6.01%
theoretical viewing angle:
horizontal °
vertical °
Relative deviation of depth displayed
theoretical depth:
20mm
Actual depth: 13.54mm, RDD: -32.3%
Image quality --maintenance of image details
-MTF D correlation coefficient
太挤了
Parameters 1 2 3
Average
2D correlation
coefficient
0.0663
0.0651
0.0652
0.0655
2D correlation coefficient after logarithmic scaling
0.639
0.636
0.638

11. Results Evaluation results based on 3D medical image Viewing angle
Parameters
1/°
2/°
3/°
Average/°
Error
horizontal
viewing angle
17.0
16.0
19.0
17.3
-9.99%
vertical
19.5
18.5
-14.1%
Relative deviation of depth displayed
theoretical depth:
48.68mm
Actual depth: 40.14mm, RDD: -17.5%
Image quality --maintenance of image details
-MTF D correlation coefficient
0.503(after logarithmic scaling)

12. 3D autostereoscopic medical display
Discussion
Error:
IP display technology
Viewing angle
lens array, pixels under lens
Displayed depth
lens array, gap
Image quality
PRR algorithm, lens array
System error
Image acquisition system
positional accuracy, environment
3D autostereoscopic medical display
platform
stereo-camera
评估方式 补充实验
评估的是显示器本身 讨论标准的误差 以及显示器本身的误差源
Evaluation results --corresponding to characteristics of IP

13. 3D auto-stereoscopic displayer
Conclusion
A standard 3D quantitative evaluation system
3D evaluation reference pattern
algorithms for quantitative analysis
IP based 3D autostereoscopic display
three parameters
viewing angle, RDD and image quality
stereo-camera
3D auto-stereoscopic displayer
workstation
platform
auto-control
image
processing
image acquisition

14. Conclusion Future work: algorithms settings of devices
3D evaluation reference patterns
algorithms
settings of devices
3D quantitative evaluation system
systemic, objective, quantitative
beneficial for design of 3D display
IP based display & other 3D autosterescopic display
评估方式

15. Thank you! http://at3d.med.tsinghua.edu.cn/
Acknowledgment: The authors thank the support of National Natural Science Foundation of China (Grant No , , ) and Grant-in-Aid of Project 985.
实验室网页链接

16. Appendix Zone plate pattern Where km denotes spatial frequency and
rm determines the maximum of the radius of the diffraction ring.
g0 is the value of the amplitude modulation.
|x| is the radius of the displayed object.
g(x) is the intensity of the generated pattern.

17. Appendix
Theoretical viewing angle of IP based 3D display

18. 2D correlation coefficient(log)
Appendix
Deviations of three factors between 3D medical images and 3D evaluation reference pattern
Parameters
Viewing angle
RDD
2D correlation coefficient(log)
Horizontal
Vertical
Medical images
17.3°
19.0°
-17.5%
0.503
Reference pattern
17.0°
20.8°
-32.3%
0.638
Relative deviation
1.7%
-8.7%
-45.8%
21.2%