[ptx] libsift control points

[Ian Sydenham]

> On Fri, 2004-10-08 at 17:07, Eric Engle wrote:
> 
>>So I backed up and took a closer look at the control points.  I did a
>>pixel-level comparison between features present in both images and found
>>that the control points were NEAR the same features but were between 5
>>and 15 pixels translated.  Unfortunately, the degree and direction of
>>translation is different at each control point.
> 
>Ed Halley wrote: 
> "What we've got here is a target-rich environment."
> 
> I can only speculate here, not having seen the images, but I think the
> algorithm is finding a TON of potential sift-able features which all
> look basically the same

Sounds like an interesting problem. I have a couple of suggestions on how to 
attempt to get a good match with images like this. I'm offering advice from a 
perspective of ignorance having not tried it, but if you are searching for some 
ideas try these:
1. Try using a small(er) mindim parameter for generatekeys when creating the 
keyfiles. This effectively creates keys for an image with smaller resolution 
which may mask out some of the detail which is finding multiple features that 
look the same. The beauty of the mindim parameter is that (although a smaller 
resolution is used to create the keys) the full resolution is still used when 
creating the actual pixel locations x and y for the control point pairs
2. If you are running windows try using autopano from http://autopano.kolor.com
This has a few parameters that do not have an equivalent in autopano-sift. The 
one that looks to be interesting is the /search: parameter which gives a few 
options for keypoint pair search.

Ian Sydenham