Skip to content


Subversion checkout URL

You can clone with
Download ZIP


Homography alignment and D_sub search window limits #23

dshean opened this Issue · 0 comments

1 participant


Still having problems with the homography alignment and D_sub values for unmapped images, even after the spheroid change. I'm seeing good, continuous disparity values in the interior, with bad values near edges/corners, usually on the negative end of the disparity range. The union of these bad values from D_sub H and V propagates to stereo_corr.

The actual alignment appears ok in most cases - not sure if there's much room for improvement there. D_sub H and V often have a "saddle" appearance when alignment isn't great, with bad disparity values in the corners. The issue in these cases may be related to the nonuniform distribution of interest points (clustering near center) used to define the transform. Regardless of the cause, the search window limits used to generate D_sub are insufficient to capture the larger disparity values near corners, and/or there may be an integer rounding issue.

I don't think I can post attachments on github. Emailing screenshot to illustrate...

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.