Depth from monocular cues – comparison and combinations

Master's Thesis, Aktuelles, Bachelor's Thesis, Ulrich Kirchmaier, Forschungspraxis, W00CGN-newscat-lkn |


Depth from Mono, or Shape from X, is a vibrant field in computer vision, which aims to estimate the underlying 3D structure of a scene on the basis of a single image. Various monocular cues can be utilized to this problem, all of which have different drawbacks and benefits depending on the image content.

 

Four widely used cues are

 

 

  • Depth from (de)focus
  • Depth from shading
  • Depth from texture
  • Depth from geometry

 

 

In this thesis, the most promising monocular depth cues are to be implemented (C++), evaluated and compared. In a second step, benefits of combinations of these cues and methods for selecting promising cues are to be examined.