ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | Q&A
Ask Your Question

How to improve terrible stereo_image_proc quality

asked 2015-05-12 02:22:37 -0500

Cerin gravatar image

Are there any other documented techniques for improving the quality of the disparity image generated by stereo_image_proc aside from what's documented in the wiki?

Following the tutorial, I spent several hours tweaking the parameters, but I found it nearly impossible to get any kind of decent quality. Any kind of surface with uniform texture either appeared as flat gray or as a constantly changing jumble of colors. Objects with a distinct texture came out the best, but even they were still barely recognizable.

image description

image description

image description

image description

Is this the best quality I should reasonably expect from a stereo camera using cheap USB webcams, or is there any way I can improve the quality?

edit retag flag offensive close merge delete



stereo vision has a history of having trouble with untextured surfaces.

ahendrix gravatar image ahendrix  ( 2015-05-12 02:59:52 -0500 )edit

1 Answer

Sort by ยป oldest newest most voted

answered 2015-05-12 02:58:16 -0500

Wolf gravatar image

Uniform texture less surfaces cannot be used for stereo vision because the basic principle of stereo vision is to match image segments captured with both cameras from the object to calculate the disparity. This requires matchable textures!

I think the bottom result is yet quite good if you use a cheap usb comsumer camera! Shaded regions, or not matchable parts in your image is something you will always have to deal with when analysing stereo images...

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools

1 follower


Asked: 2015-05-12 02:22:37 -0500

Seen: 308 times

Last updated: May 12 '15