The dense matcher code has been used quite a bit but not in the last year or so.
I will check to see what the current state of the program is and I will report back.
-----Original Message-----
From: Geoffrey Cross []
Sent: Monday, September 16, 2002 6:42 AM
To: 'Gregory Hassett';
Cc: Tu, Peter (Research)
Subject: RE: [Vxl-users] Question regarding dense_matcher_full


I haven’t tried this, but are you using colour images?  If so, try greyscale pgms (bmp loading was broken last time I tried, although it should be fixed now…): it might be that this is assumed somewhere.


Peter Tu wrote the original code: perhaps he can help out…  I’ve seen the code running in the past, and it appears to be pretty good.





-----Original Message-----
From: [] On Behalf Of Gregory Hassett
Monday, September 16, 2002 11:21 AM
Subject: [Vxl-users] Question regarding dense_matcher_full




I am trying to get bootstrapped with VXL – I have built the code on Windows XP and am now beginning to experiment.  My starting point is the dense_matcher_full application which closely approximates the task that I am trying to accomplish with VXL.  However, it assert’s in a way that is quite puzzling, as I cannot see how the program could have ever run without this assertion firing.


I know this is a bit obtuse; but the bottom line is that I can’t see how this program (dense_matcher_full) could ever work under any circumstances.

Has anyone gotten this app to successfully generate a disparity image?  Here’s the obtuse details:


I start this example from the command line like this:


         dense_matcher_full left.bmp right.bmp out.bmp


Stepping through the source shows me that the code attempts to call the functions



vsrl_stereo_dense_matcher::set_correlation_range (10);




But the first thing that takes place inside ::execute is a call to vsrl_image_correlation::initial_calculations() which calls compute_correlation() which in turn will assert quite obviously (at vrsl_image_correlation.cxx, line 160) if the correlation_range_ member of the vrsl_image_correlation object is greater than zero.


// thanks // greg //