Items with no label
3335 Discussions

Visible image and infrared image registration

idata
Employee
2,523 Views

I got visible and infrared images of the same resolution using sr300, but I found that the two images have different ranges, and the target has a different image ratio. I want to merge the two images. How can I proceed?

0 Kudos
9 Replies
MartyG
Honored Contributor III
1,268 Views

RealSense cameras align between depth and some other stream. So I do not believe you would be able to align IR to color. It would have to be depth to color.

In the '2016 R2' SDK, you may be able to do such an operation with the script in the link below:

https://software.intel.com/sites/landingpage/realsense/camera-sdk/v1.1/documentation/html/index.html?doc_essential_aligned_capturing.html Intel® RealSense™ SDK 2016 R2 Documentation

In the current RealSense SDK 2.0, which is compatible with SR300, there is an Align sample program.

https://github.com/IntelRealSense/librealsense/tree/master/examples/align librealsense/examples/align at master · IntelRealSense/librealsense · GitHub

When the SR300 is used with RealSense SDK 2.0, depth images are always pixel-aligned with IR images, and the depth and IR images have identical intrinsics.

https://github.com/IntelRealSense/librealsense/wiki/Projection-in-RealSense-SDK-2.0# sr300 Projection in RealSense SDK 2.0 · IntelRealSense/librealsense Wiki · GitHub

0 Kudos
idata
Employee
1,268 Views

So you mean that we can't align color and infrared images?

Do you know how to get aligned? Thank you!

0 Kudos
MartyG
Honored Contributor III
1,267 Views

Although alignment to depth is the method supported by RealSense, a member of Intel's GitHub support team for RealSense offered an approach to try if you really need to find a way to align IR to color.

https://github.com/IntelRealSense/librealsense/issues/1556# issuecomment-389061477 Align infrared frame to color · Issue # 1556 · IntelRealSense/librealsense · GitHub

0 Kudos
idata
Employee
1,267 Views

Do you know other methods? This seems to be a bit problematic. Thank you very much!

0 Kudos
MartyG
Honored Contributor III
1,268 Views

I researched your question extensively, but everyone everywhere agrees that aligning to depth rather than IR is the appropriate method. There was no alternative to the workaround suggested by Intel's support member that could be found.

0 Kudos
idata
Employee
1,268 Views
0 Kudos
GYuen
Beginner
1,267 Views

Hi Marty

 

We need to align infrared to RGB in the SR305, but found the size of image for ir is much smaller (although both at 640x480), while the depth is same as the RGB,

this means the infrared is certainly not aligned pixel by pixel (it is also offset, see attached), any tips would be much appreciated.

 

We tried using the difference in POV for ir and RGB and do inverse trigonometry this gets the sizes closer, but we only have the POV for 1920x1080, what POV I should use for the RGB camera for 640x480 ?

 

thanks a lot,

Geoff

0 Kudos
MartyG
Honored Contributor III
1,267 Views

@GYuen​ Usually, depth is aligned to color or to IR, but depth must be one of the stream types in alignment (and there is not support in the SDK for aligning color to IR).

 

I hope that the link below will be helpful to you, as it aligns all three stream types.

 

https://github.com/IntelRealSense/librealsense/issues/5093

0 Kudos
GYuen
Beginner
1,264 Views

Hi Marty

Are you aware of anyone using the SR300/SR305 to 'read' the palm/hand print for identification ? 

Could be using NIR ...

thanks

Geoff

0 Kudos
Reply