单幅图像深度重建:基于单眼线索的深度重建Shape From Shading

本章介绍单幅图像深度重建中基于单眼线索的深度重建方法,包括从明暗(Shading)、灭点(Vanishing Point)、散焦(Defocus)、纹理(Texture)四个角度,第一节介绍了明暗的相关概念以及数学模型、实现方法。
展开查看详情

1.第九章 单幅图像深度重建 Depthmap Reconstruction Based on Monocular cues

2.深度图

3. 章节安排  基于单眼线索的深度重建  Shape From Shading  Shape From Vanishing Point  Shape From Defocus  Shape From Texture

4.Shape From Shading

5. What is Shading?  Well… not shadow…  We can’t reconstruct shape from one shadow…

6. What is Shading?  Variable levels of darkness  Gives a cue for the actual 3D shape  There is a relation between intensity and shape

7. Shading Examples  These circles differ only in grayscale intensity  Intensities give a strong “feeling” of scene structure

8.What determines scene radiance? n

9.

10. Surface Normal Convenient notation for surface orientation A smooth surface has a tangent plane at every point We can model the surface using the normal at every point

11. The Shape From Shading Problem  Given a grayscale image  And albedo  And light source direction  Reconstruct scene geometry  Can be modeled by surface normals

12. Lambertian Surface  Appears equally bright from all viewing directions  Reflects all light without absorbing  Matte surface, no “shiny” spots  Brightness of the surface as seen from camera is linearly correlated to the amount of light falling on the surface Here we will discuss only n Lambertian surfaces under point-source illumination

13. Some Notations: Surface Orientation

14. Some Notations: Surface Orientation

15.Reflectance Map

16. Reflectance Map • Lambertian case I cos  i n s   pps  qqs 1 R  p, q  2 2 2 2 p  q 1 p  q 1 S S Reflectance Map Iso-brightness contour (Lambertian) cone of constant  i

17. Reflectance Map • Lambertian case iso-brightness contour q 0.8 0.9 1.0 R  p, q  0.7  pS , q S  p  i 90  ppS  qqS 1 0 0.3 0.0 Note: R p, q  is maximum when  p, q   pS , qS 

18. Reflectance Map Example • Brightness as a function of surface orientation Lambertian iso-brightness contour q surface 0 .8 0.9 1.0 R p, q  0.7  pS , qS  p  i 90  ppS  qqS  1 0 0.3 0.0

19. Reflectance Map of a Glossy Surface  Brightness as a function of surface orientation Surface with diffuse and glossy components

20. Reflectance Map Examples  Brightness as a function of surface orientation

21.Graphics with a 3D Feel

22.Shape From Shading? q p

23. Shape From Shading!  Use more images  Photometric stereo  Shape from shading  Introduce constraints  Solve locally  Linearize problem

24. Photometric Stereo  Take several pictures of same object under same viewpoint with different lighting q p 1 S , q1S  p

25. Photometric Stereo  Take several pictures of same object under same viewpoint with different lighting q p 1 S , q1S  p p 2 S 2 , qS 

26. Photometric Stereo  Take several pictures of same object under same viewpoint with different lighting q p 1 S , q1S  p p 2 S 2 , qS  p 3 S , q 3S 

27. Photometric Stereo Lambertian case:   kc  I  kc cos  i n s  1    n Image irradiance: s2 s3 I1 n s1 s1 v I 2 n s 2 I 3 n s 3 • We can write this in matrix form: I  1  s T   I    s T  n 1  2  2  I 2   T s 3 

28.改变光源所获得的同一个球的五幅图像

29.g  x, y 