Ask Your Question

Is it possible to apply superpixel algorithm on depth data?

asked 2017-07-20 03:47:11 -0500

vps gravatar image

Hi All, I am currently working on depth image processing in openv with c++.. Is it possible to apply the superpixel algorithm on depth data? Thank you.

edit retag flag offensive close merge delete


should be technically possible,but be careful there: while e.g. SLIC supports 8u 8s 16u 16s 32s 32f , SEEDS only has 8u 16u 32f (you might have to adapt your data)

how do you get your depth image ? (any noise will be harmful here)

berak gravatar imageberak ( 2017-07-20 06:56:48 -0500 )edit

Hi Berak, Thank you for your response. I am getting the depth image from stereo camera. Do you any references for it, as I am unable to find for depth data.

vps gravatar imagevps ( 2017-07-20 07:49:58 -0500 )edit

not really, been digging through the code, but forgot to take liinks, unfortunately ;(

berak gravatar imageberak ( 2017-07-20 11:48:51 -0500 )edit

no problem berak :)

vps gravatar imagevps ( 2017-07-25 02:18:49 -0500 )edit

1 answer

Sort by ยป oldest newest most voted

answered 2017-07-24 05:02:05 -0500

arrfou gravatar image

Well, Superpixels extractor algorithms are usually applied to Color image. For example SLIC algorithm uses CIElab color space to which pixel should be assigned together. Thus, If you want to apply directly SLIC to depth image which has only depth information, the simple answer is NO, otherwise you have to modify the SLIC Kernel. However, in 3D Superpixels are called Supervoxels thus you have to look for Supervoxel methods (Check PCL library fro VCCS).

If you need more information do not hesitate to ask.

edit flag offensive delete link more


arrfou .. I am very thankful about your answer. I am using c++ language with opencv library. I am working on human skeleton extraction from depth data only. Do you know some research papers or methods for it?

vps gravatar imagevps ( 2017-07-25 02:18:22 -0500 )edit

Hello, I am not quit familiar with human skeleton extraction methods. I am more on 3D segmentation and stereo-vision area.

arrfou gravatar imagearrfou ( 2017-07-25 06:31:42 -0500 )edit

Thank You. I am capturing the input from stereo camera only. Is it possible you to suggest 3D segmentation method for depth data?

vps gravatar imagevps ( 2017-07-25 08:57:10 -0500 )edit

Question Tools

1 follower


Asked: 2017-07-20 03:47:11 -0500

Seen: 270 times

Last updated: Jul 24 '17