2

Texture is very helpful for stereo matching. However in real environment, untextured areas always exit. Consistent semi-global matching is proposed to deal with untextured regions. I start to read the paper Stereo Vision in Structured Environments by Consistent Semi-Global Matching. Fixed bandwidth Mean Shift segmentation is used.

If some have some experience with Consistent Semi-Global Matching, I hope to learn whether the algorithm is good for untextured regions. How about the complexity of the algorithm? Is there any better way to deal with untextured regions?

Jogging Song
  • 273
  • 2
  • 10

1 Answers1

1

I don't know about stereo matching but there are some "direct" monocular odometry/SLAM libraries that do not use common image features but rather whole image edges and gradients and are said to need very little texture. Monocular odometry basically also performs triangulation like stereo vision. See my post here: I mean especially LSD-SLAM and DSO. There may also be a library that uses "direct" approach for stereo vision (I dont know any myself though).

Kozuch
  • 643
  • 1
  • 7
  • 18