search

bupt-ai-cz / LLVIP

LLVIP: A Visible-infrared Paired Dataset for Low-light Vision
610 stars 65 forks source link

About registration methods. #17

Closed socome closed 2 years ago

socome commented 2 years ago

Thank you for your great work!

I have some question about your paper.

  1. In your paper, you said "We first manually select several pairs of points that need to be aligned between the two images, then calculate the projection transformation to deform the infrared image". Is it means that you calculated the Homography matrix between visible and infrared images..? (if it is corrected, could you provide the value of homography matrix..?)

  2. When i saw the some of image pairs, i think that infrared images are only resized. Do you crop the infrared images when you cut out the images to get the registered image pairs..?

Looking forward to your reply.

SantJay commented 2 years ago

  1. Yes. We implemented the projection transformation using Matlab's cpselect function, which returns the Homography matrix. We use the same transformation matrix for images in one scene(e.g. 01, 02,...). Here is the transformation matrix for each scene (scene 01 is missing): transformationMatrix.zip

  2. After the registration, the infrared images have the same size as the visible images, and we cropped the images of the two domains the same.

socome commented 2 years ago

@SantJay Thank you! :)